# How to Write a Corporate AI Training Technical Specification (RFP)? (With Template) > Source: https://sukruyusufkaya.com/en/blog/kurumsal-ai-egitimi-teknik-sartnamesi > Updated: 2026-07-09T17:41:28.533Z > Type: blog > Category: yapay-zeka **TLDR:** How do you write an AI training technical specification (RFP)? Audience, scope, learning outcomes, evaluation criteria, KVKK, a scoring table, and a copy-ready template in this comprehensive procurement guide. How do you write an AI training technical specification (RFP)? An AI training technical specification is a formal document (RFP) that defines, item by item, what an organization expects from suppliers when procuring AI training; a sound specification includes twelve components measurably, from purpose and scope to learning outcomes, from instructor qualifications to KVKK/GDPR. This specification turns an intangible service that cannot be seen before the contract into a concrete request that suppliers bid on the same basis and the organization can compare objectively. Many managers who buy corporate training experience the same disappointment: the supplier chosen with a brilliant presentation fails to create the expected behavior change when the training ends. The root cause is almost always the same — the request was not written clearly enough from the start. A well-prepared AI training technical specification fills exactly this gap: it writes down in advance what will be delivered, for whom, with which outcomes, measured how, and under which conditions. This guide covers, with the rigor of a management consultant, every step of preparing a strong AI training technical specification from scratch: the purpose and importance of the specification, an item-by-item breakdown of the twelve components, a supplier evaluation scoring table, a copy-ready example template, the Türkiye and KVKK context, role/sector examples, an implementation checklist, common mistakes, and an end-to-end procurement process. ## Why Is an AI Training Technical Specification So Important? AI training is one of the services organizations buy the most today but define the worst. When buying a software license, you compare its features item by item; when buying hardware, you look at its technical specs. Yet training is a service whose "quality cannot be seen" until you sign the contract, and even until the training ends. This invisibility is exactly what makes an AI training technical specification critical: the specification manages this uncertainty by making the expectation concrete before the purchase. The first reason is comparability. When you ask three suppliers for proposals without a specification, each sends offers in its own format, highlighting the aspects it emphasizes, impossible to compare with one another. One describes instructor experience, another highlights price, a third gives a long topic list. A common specification forces them all to answer the same questions in the same format; so you can compare offers apples to apples. This is the foundation of the procurement process. The second reason is accountability. In a corporate purchase, you must be able to give a document-based answer to "why did we choose this supplier?" — especially in public institutions, regulated sectors, and companies with strong internal audit. A well-written AI training technical specification and a scoring table tied to it take the decision out of personal preference and make it defensible. When evaluation criteria are defined in advance, the choice becomes an objective result, not a subjective impression. The third reason is expectation alignment. The most common dispute in training procurement begins with the sentence "this is not what we expected." While the organization expected its team to become able to produce real AI projects, the supplier may have delivered a general awareness seminar. This gap is inevitable when scope and learning outcomes are not written clearly from the start. The specification is a pre-contract alignment tool that makes both sides imagine the same result. To see what AI training means within the organization in a broader frame, the what is corporate AI training guide is a good start. The fourth and least-discussed reason is managing risk up front. A training may touch the organization's real data, process participants' personal data, produce organization-specific materials, and access strategic knowledge. These risks — KVKK compliance, intellectual property, confidentiality — must be addressed at the specification stage, not after the contract. Skipping the specification and directly accepting an offer leaves these risks invisible and unmanaged. In this sense, an AI training technical specification is not only a purchasing tool but also a risk management tool. The fifth reason is scalability and institutional memory. An organization may buy its first AI training ad hoc, without a specification; but by the second, third, and tenth training, this approach becomes unsustainable. A well-structured AI training technical specification template ensures the organization does not start from scratch with each new procurement: a template built once is reused many times by adapting it to context, and improves a little each time. This turns the procurement process from a person-dependent art into an institutionalized capability. Building the specification correctly once prepares the ground for dozens of future procurements; so the investment in the first specification produces a return far beyond a single training. The real value of an AI training technical specification is not the PDF it produces but the thinking discipline it forces on the organization to produce that PDF: "What do we actually expect from this training? Whose behavior, and which behavior, will change? How will we measure success?" Answering these questions while writing the specification designs the training far more accurately from the start. ## What Are the Components of an AI Training Technical Specification? A sound AI training technical specification consists of twelve components. Each of these components closes a specific uncertainty in the procurement process; when one is missing, that gap turns into either subjective interpretation or a later dispute. Below we first take a bird's-eye view of all the components, then open each in its own heading item by item. These twelve components form a chain that feeds one another: purpose determines scope, scope determines curriculum, curriculum determines learning outcomes, outcomes determine assessment, and assessment determines acceptance criteria. If one link of the chain is weak, all of it weakens. That is why the specification should be written not piece by piece but with a holistic logic. Now let us address each component in detail. ## How Are Audience and Prerequisites Defined? The success of a training often depends, before its content, on being delivered to the right audience at the right level. That is why the definition of audience and prerequisites is one of the most decisive components of an AI training technical specification. The same "AI training" title means strategic awareness for a senior executive, practical tool use for an analyst, and technical depth for a developer. Asking for content without defining the audience is like a tailor sewing a suit without taking measurements. The audience definition should be done in three dimensions. The first is role and function: is the training aimed at executives, middle management, technical teams, or all employees? Each role's need and attention span differ. The second is current knowledge level: are participants completely new to AI, familiar with basic concepts, or advanced? This level can be expressed more clearly with the what is AI literacy framework. The third is group size and heterogeneity: a homogeneous expert group of 10 and a mixed audience of 200 require very different pedagogical approaches. Prerequisites are the minimum knowledge, skills, or tool access participants must have before coming to the training. For example, in a hands-on generative AI training, access to a laptop and certain tools may be a prerequisite; in an advanced training, mastery of basic concepts (what is AI, what is machine learning) may be expected. Stating prerequisites in the specification lets both the supplier calibrate the content correctly and the organization prepare participants in advance. The most common audience mistake is gathering people of very different levels into a single training. Advanced participants get bored, beginners cannot keep up; in the end no one gets full value. In the specification, either define homogeneous groups or ask the supplier for a level-differentiated approach. Making this distinction from the start saves half the training. A practical way to define the audience correctly is the persona approach: a concrete profile like "Ayşe, marketing manager, has heard of AI but not used it, can spare 20 minutes a day, her goal is to speed up her team's content production" tells the supplier far more than the abstract phrase "middle manager." Defining two or three representative personas in the specification ensures the content is designed for real people. ## How Are Scope and Curriculum Defined? The scope definition is the backbone component of an AI training technical specification; because it determines what the training includes and — equally importantly — what it does not include. A scope left vague creates a two-way problem: the supplier either covers too little and stays superficial, or promises everything and cannot cover any of it in depth. A clear scope definition protects against both extremes. A good scope definition answers three questions. The first is thematic boundaries: will the training be general AI awareness, or focus on a specific subdomain like generative AI tools, data literacy, or AI governance? The second is depth level: will each topic be introduced conceptually or covered hands-on? The third is what is out of scope: saying in the specification "this training does not cover the following" is the strongest tool of expectation alignment. To set up the technical details of scope correctly, the corporate AI training program selection guide, which discusses which program type is right within the organization, is complementary. The curriculum is the scope opened up module by module. There are two approaches when asking for a curriculum in the specification. In the strict approach, the organization dictates modules and topics one by one; this is good if the need is very clear but may prevent benefiting from the supplier's expertise. In the flexible approach, the organization gives the goals and learning outcomes, leaves the curriculum design to the supplier, and compares offers on curriculum quality. In most mature procurements the flexible approach is preferred, because it reveals a good supplier's pedagogical design value-add. The critical point when asking for a curriculum is the difference between a "topic list" and a "learning journey." A weak curriculum is a list of disconnected topic headings: "1. History of AI, 2. Machine learning, 3. Deep learning..." A strong curriculum carries a progression logic: each module builds on the previous one and moves the participant step by step toward a competency. Asking the supplier in the specification to explain not only the topics but the logical flow between them and which learning outcome each module serves makes curriculum quality visible. ## How Are Learning Outcomes Written? Learning outcomes are the heart of an AI training technical specification; because they form the only objective basis for measuring the training's success. A learning outcome defines what the participant will be able to do at the end of the training — note, not "what they will learn," but "what they will be able to do." This distinction is critical: "learning AI" is unmeasurable, but "producing three AI use cases for one's own work" is measurable. The classic tool for writing measurable learning outcomes is Bloom's taxonomy. This framework orders cognitive skills in a hierarchy and proposes observable verbs for each level: remembering (defines, lists), understanding (explains, exemplifies), applying (uses, applies, computes), analyzing (compares, distinguishes), evaluating (critiques, selects, justifies), and creating (designs, produces, develops). Every learning outcome in the specification should begin with one of these verbs — because verbs like "understands" or "knows" are unmeasurable, while verbs like "applies" or "produces" define a testable behavior. Three practical rules help when writing learning outcomes. First, each outcome should have a business context: not "writes a prompt," but "writes a prompt for a task in their own department." Business context ties learning from an abstract skill to organizational value. To deepen the prompt concept itself, the what is prompt engineering guide can be recommended to participants. Second, the number of outcomes should be reasonable: promising 20 learning outcomes in a one-day training is unrealistic; 4-6 well-defined outcomes are worth more than 20 vague ones. Third, each outcome should be tie-able to an assessment method: if you cannot think of how to measure an outcome, that outcome is probably not concrete enough. The real power of learning outcomes in the specification is the bond they form with the later components. Assessment criteria are designed according to these outcomes; acceptance criteria are tied to the realization of these outcomes; even the curriculum is shaped to serve these outcomes. That is why writing learning outcomes is actually building the spine of the whole specification. Well-written outcomes turn the question "was the training successful?" from a matter of debate into a matter of measurement. ## How Are Duration, Format, Materials, and Platform Stated in the Specification? As much as a training's content, how it is delivered determines its success; that is why duration, format, materials, and platform should be addressed separately and clearly in an AI training technical specification. These components matter for pedagogical value as well as logistics and cost; when left vague, offers become incomparable and surprises arise in practice. **Duration** defines how many hours/days the training will be and how this time will be distributed. The critical distinction here is between an "intensive single block" and "spread-out sessions." A one-day intensive training works for awareness but is usually insufficient for behavior change; because learning requires repetition and practice. A program spread over weeks, with practice tasks in between, is more effective for lasting competency. In the specification, you should state not only the total duration but also how you want it structured. **Format** defines whether the training will be in-person, online (synchronous/asynchronous), or blended (hybrid). Each format has an advantage and a limit: in-person provides interaction and focus but has high logistical cost; online is scalable and flexible but requires attendance discipline; blended balances the two. Stating the format preference and its rationale in the specification lets the supplier bid correctly. **Materials and platform** are the invisible factor determining the training's durability. Materials cover slides, exercise workbooks, sample datasets, recorded videos, and reference documents. The platform is the learning management system (LMS) where these materials are delivered and attendance is tracked. In the specification, you should clarify whether materials will be given to participants, how long they will be accessible, and whether platform access is included. A participant who cannot access materials after the training forgets most of what they learned within a few weeks; that is why reinforcement material is the key to preserving the training's real value. A point to watch especially with materials is currency. Because the AI field changes rapidly, a material prepared a year ago may be outdated today. Requiring in the specification that materials reflect current tools and approaches, and if possible asking when the material was last updated, keeps the content fresh. This is critical especially in areas that change monthly, like generative AI. On the platform side, it matters that the learning management system (LMS) can not only host materials but also track attendance and progress. If your acceptance criteria include an attendance rate threshold, you need a platform that measures this rate; otherwise the criterion stays on paper. Asking in the specification what data the platform can report (attendance, completion, assessment results) ties the assessment component to the materials-platform component. Also, if the platform needs integration with internal systems (e.g., an existing corporate LMS), this technical requirement should be stated in the specification; an integration need that surfaces later disrupts both cost and schedule. ## How Are Assessment and Acceptance Criteria Built? You can only know whether a training truly worked by measuring it; that is why assessment is one of the most neglected but most valuable components of an AI training technical specification. Without assessment, when a training is complete you can only say "it was done"; you cannot say "it worked." The specification closes this gap by defining the assessment method in advance. The classic framework for training assessment is the Kirkpatrick model. This model measures training impact at four levels. **Level 1 — Reaction**: how participants found the training (satisfaction survey). **Level 2 — Learning**: what participants learned (pre-test/post-test, quiz). **Level 3 — Behavior**: what participants changed in their work (post-training observation, manager assessment). **Level 4 — Results**: what the training added to business results (efficiency, quality, business metrics). Most organizations stop at Level 1 — they run a satisfaction survey and stop there; yet the real value is at Levels 3 and 4. Asking the supplier in the specification which levels it will assess sets the seriousness of the training from the start. The most critical extension of assessment in the specification is the acceptance criteria. Acceptance criteria are the measurable thresholds that must be met for the training to count as "completed and payable." An unmeasurable acceptance criterion — for example "the training must be successful" — is useless; because who defines success, and how, is unclear. Good acceptance criteria are numerical and written in advance: a minimum attendance rate (e.g., at least 80% of participants attending sessions), an assessment pass threshold (a given proportion of participants passing the post-test), a satisfaction score (a certain average out of 5), and concrete output delivery (each participant producing an implementation plan). Tying acceptance criteria to the payment condition in the specification is the strongest mechanism to make the supplier focus on the result. For example, tying part of the payment to the Level 2 assessment passing a certain threshold motivates the supplier not merely to "deliver the training" but to "ensure learning." This is a simple but effective procurement technique that aligns interests. A warning is also in order when designing assessment: assessment exists not to punish the participant but to verify and improve learning. An overly hard or threatening assessment can put participants off the training. The ideal assessment both proves learning and offers the participant feedback showing their own progress. Stating this constructive purpose of assessment in the specification lets the supplier design an assessment in the right tone. Assessment in AI training has a special difficulty: most of the skill learned cannot be measured with a multiple-choice test. You can only understand whether a participant has truly become able to integrate AI into their workflow through an application assignment — for example, designing and presenting an AI use case for their own department. That is why the most valuable form of assessment in advanced AI trainings is not a knowledge test but a hands-on project. Requiring in the specification a "concrete output" (project, plan, application example) among the assessment criteria is the most effective way to guarantee the transfer of learning to the real world. This output also gives the organization concrete evidence of the training; the use cases participants produce often become a more valuable output than the training itself. Another dimension of assessment is its timing. An end-of-training assessment (post-test) measures learning but cannot measure behavior change; because behavior can only be observed weeks after the participant returns to work. That is why a mature assessment design includes a "follow-up assessment" one or two months after the training: are participants actually using what they learned? Requesting this follow-up assessment in the specification ties the training from a momentary event to a lasting behavior change and holds the supplier accountable for that durability too. ## Why Are Instructor Qualifications and the Reference Requirement Critical? The quality of a training depends largely on the quality of the instructor; that is why instructor qualifications and the reference requirement should be defined meticulously in an AI training technical specification. Even the best curriculum produces no value if the instructor delivering it is not competent. This is especially true in AI training, because the field requires both technical depth and real-world application. Instructor qualifications should be defined with two types of criteria. **Formal criteria**: educational background, relevant certifications, academic or professional titles. These provide a floor but are not sufficient alone; because a certificate is not a guarantee of good teaching. **Evidence-based criteria**: experience working on real AI projects, similar trainings delivered before, domain expertise, and concrete evidence of teaching skill. In AI training, the most valuable instructor is the one who knows concepts not only from books but from having applied AI in a production environment; because what participants value most is real-world examples. To understand the quality of this experience, the what is AI consulting guide shows how consulting and training feed each other. There are practical ways to secure instructor qualifications in the specification. First, asking the supplier to submit the instructor's CV in advance. Second, if possible, requesting a short teaching sample or demo — because a person being knowledgeable does not mean they can teach well. Third and most importantly, stipulating that instructor changes can be made only with the organization's approval. A common procurement trap is showing an experienced instructor in the offer and then handing the training to someone else; this "bait-and-switch" tactic must be prevented with a specification clause. **The reference requirement** is proof of the competence the supplier claims. References are concrete examples showing the supplier has delivered similar trainings to similar organizations before. When asking for references in the specification, watch three things: the reference's similarity (same sector, same scale, same topic), currency (within the last two-three years), and verifiability (a person or organization that can be contacted). Skipping the reference requirement means accepting an unproven competence claim; this is one of the most expensive mistakes in training procurement. A classic disappointment in training procurement is that the impressive expert who gave the offer presentation never appears in the actual training. Be sure to add this clause to your specification: "The instructor(s) stated in the offer will deliver the training personally; a change can be made only with the organization's written approval and with an instructor of at least equivalent qualification." This single sentence prevents the most common procurement deception. ## How Are the Pricing Structure and Delivery/Logistics Defined? Pricing is an important part of the procurement decision but should not be the only determinant; that is why in an AI training technical specification you should ask for price not as a single number but as a comparable structure. Offers taken without standardizing the pricing structure are like comparing apples to oranges: one offers an all-inclusive package, another gives a low base price and inflates it with extras. The first step to clarifying the pricing structure is to determine the pricing model. There are two basic models. **Fixed package price**: a total fee for a given group size and scope; it provides budget predictability and is usually advantageous for small-to-medium groups. **Per-participant price**: a unit fee per participant; it offers scalability when the number of participants is uncertain or very large. Asking for both models comparably in the specification lets the organization choose the most suitable one for its situation. To see the corporate context of pricing more broadly, the AI consulting prices and, for budget planning, the corporate AI budget planning guides help. The most important discipline in pricing is to clearly list what the price covers. In the specification you should ask the supplier to state whether the following are included in the price: training materials, platform/LMS access, certification, assessment tools, instructor transport and accommodation costs, post-training support, and material updates. Leaving no hidden cost both prevents budget surprises and makes offers truly comparable. The clause "every item not included in the price must be stated separately" makes the pricing section of the specification transparent. **Delivery and logistics** define how the training will practically come to life: training venue (on-site, supplier venue, online), schedule and dates, session plan, organization of participant groups, technical requirements, and the communication/coordination process. Logistical details may seem trivial but in practice directly affect the training's success: a mistimed training (e.g., during a busy business period) lowers attendance; an online session without prepared technical infrastructure loses the first half hour. Clarifying in the specification which side holds the logistical responsibilities (the organization or the supplier) prevents coordination chaos in the implementation phase. ## How Are the Intellectual Property, Confidentiality, and KVKK Clauses Written? AI training can touch the organization's most sensitive assets: personal data, organization-specific knowledge, and produced intellectual property. That is why the intellectual property, confidentiality, and KVKK clauses are too critical to leave to after the contract in an AI training technical specification; they must be written from the start. Skipping these clauses creates both legal risk and enterprise-knowledge leakage risk. **Intellectual property (copyright)** defines the ownership of materials and content produced during the training. Two questions must be clarified here. First, the ownership of standard training material usually stays with the supplier but a usage license is granted to the organization; the scope of this license (how many people, for how long, reuse rights) should be set in the specification. Second and more importantly, who owns the content produced specifically for the organization during the training (e.g., examples adapted to the organization's own use cases, projects produced by participants)? Stipulating in the specification that content produced specifically for the organization will belong to the organization prevents copyright disputes that may arise later. **Confidentiality** defines the protection of the organizational knowledge the supplier accesses during the training. An instructor may learn the organization's internal processes, strategic priorities, and even weaknesses during the training. Stipulating a confidentiality undertaking (NDA) in the specification and stating that this undertaking will continue after the training protects organizational knowledge. Especially in competitive sectors, the possibility of the same instructor also training a competitor makes the confidentiality clause even more important. **The KVKK (Personal Data Protection Law)** clause is one of the most critical components in the Türkiye context. The training process inevitably processes personal data: participant name, email, assessment results, and sometimes internal sample data. In the specification you should ask the supplier for these commitments: processing data only for training purposes, deletion or return after training, restrictions on transfer to subcontractors, appropriate data security measures, and the signing of a data processing agreement (DPA). You can find the basics of KVKK in what is KVKK and the definition of personal data in what is personal data. The most dangerous scenario in hands-on AI trainings is moving the organization's real data (actual customer records, actual financial data) into the training environment. This creates KVKK violation and data leakage risk. In the specification, if real data will be used in training, be sure to add an anonymization or synthetic data requirement. We cover how data anonymization is done in what is data anonymization. If the training is for an organization operating in the European Union or processing EU citizens' data, GDPR and the increasingly effective EU AI Act come onto the agenda alongside KVKK. These regulations may bring additional obligations to AI systems and therefore to their trainings. To understand the regulatory framework, the what is the EU AI Act and, for a KVKK-compliant architecture, the what is KVKK-compliant AI guides strengthen the compliance clauses of the specification. Compliance is not a formality in training but an institutional responsibility. ## How Is a Supplier Evaluation Scoring Table Built? When the specification is sent to suppliers and offers are collected, the critical moment arrives: which supplier will be chosen? The tool that saves this decision from subjectivity is the supplier evaluation scoring table. This table gives each supplier an objective total score by weighting evaluation criteria and ensures the offer that "gets the highest score," not the one that "feels best," is chosen. The first step to building the scoring table is to gather criteria into two main blocks: technical competence and commercial offer. **Technical competence** usually includes instructor experience, curriculum fit, learning-outcome quality, methodology, references, and material quality. **Commercial offer** covers price, payment terms, and delivery flexibility. The second step is to give these two blocks a weight. A common distribution is 70/30 (technical/commercial): this shows that quality is more important than price but price is not ignored either. In critical and strategic trainings the technical weight can rise even further (e.g., 80/20); in commoditized, standard trainings the price weight can rise. In the scoring table, a score range (typically 1-5) and a weight are defined for each criterion. Each supplier's total score is the sum of the products of criterion score and weight. What matters in the table is that how scores will be given is defined in advance: for example, in the "instructor experience" criterion, what does 5 points mean, what does 3 points mean? When these definitions (the scoring rubric) are written in advance, the evaluation does not vary from person to person and stays consistent. There is a special way to score price. Price should be evaluated not directly as an amount but as a relative score: the lowest offer gets full marks, the others are scored relative to it. Thus price can be evaluated on the same scale (1-5) as technical competence. This approach makes price important but not the sole determinant. In the end, the supplier with the highest total score is chosen; and this choice provides a document-based, defensible answer to "why this supplier?" It is essential to fix the criteria and weights of the scoring table before seeing the offers. Adjusting the weights after seeing the offers turns into a manipulation that, without your realizing it, favors the supplier you prefer. Writing and locking the criteria and weights from the start protects the fairness and defensibility of the process. ## A Copy-Ready Example AI Training Specification Template, Item by Item Now let us turn all the components into a copy-ready example template you can adapt directly. The template below contains fillable headings and example phrasings for each of the twelve components. It is recommended to adapt this template to your organization's context (sector, team size, maturity level) and to replace the general phrases in brackets with your own concrete needs. **1. Purpose and Scope:** "The purpose of this specification is to procure corporate AI training in the field of [topic] for [target unit] within [organization name]. The training serves the goal of [main objective]. In scope: [items]. Out of scope: [items]." **2. Audience and Prerequisites:** "The training is aimed at approximately [number] participants in the role of [role/position]. The participants' current AI knowledge level is assumed to be [beginner/intermediate/advanced]. Prerequisites: [required knowledge/tool access]. Groups will be [homogeneous/heterogeneous]." **3. Curriculum and Content:** "The supplier will propose a curriculum serving the learning outcomes below. The curriculum should be presented not as a topic list but as a learning journey explaining the logical flow between modules and which outcome each module serves." **4. Learning Outcomes:** "At the end of the training the participant will be able to: (a) [measurable verb + business context]; (b) [measurable verb + business context]; (c) [measurable verb + business context]. Each outcome must be testable with the relevant assessment method." **5. Duration and Format:** "The training will last a total of [hours/days] and be delivered in [in-person/online/hybrid] format. The duration will be structured as [single block/spread-out sessions]. [Practice tasks] are expected between sessions." **6. Materials and Platform:** "The supplier will provide [slide/exercise workbook/video/sample data] materials. Materials will be accessible to participants via [platform/LMS] for [duration]. Materials must reflect current tools and approaches; the last update date must be stated." **7. Assessment:** "Training impact will be measured up to at least [level] of the Kirkpatrick model: [reaction survey / pre-post test / behavior observation]. Assessment tools will be provided by the supplier and results reported to the organization." **8. Instructor Qualifications:** "The instructor(s) must have at least [years] of experience in [relevant field] and hands-on experience on real AI projects. Instructor CVs will be attached to the offer. A change to the instructor stated in the offer can be made only with the organization's written approval and an instructor of at least equivalent qualification." **9. Reference Requirement:** "The supplier will present at least [number] training references of similar scope and scale carried out within the last [two-three] years. References will be given verifiably, with contact information." **10. Pricing Structure:** "The supplier will present both [fixed package] and [per-participant] pricing separately. The items included in the price (materials, platform, certification, transport, support) will be clearly listed; every item not included will be stated separately." **11. Delivery and Logistics:** "The training will be held at [place], during [date range]. The schedule, session plan, and technical requirements will be stated in the offer. Logistical responsibilities (venue, equipment, coordination) will be covered by [party]." **12. Intellectual Property, Confidentiality, and KVKK:** "Content produced specifically for the organization during the training belongs to the organization. The supplier will keep confidential all organizational knowledge accessed during the training and will sign a confidentiality undertaking (NDA). Personal data will be processed only for training purposes and will be [deleted/returned] after training; the parties will sign a data processing agreement. If real data is used in training, anonymized or synthetic data will be used." **Acceptance Criteria:** "The training is deemed accepted when the following measurable criteria are met: (a) at least [%] of participants attend the sessions; (b) at least [%] of participants pass the final assessment; (c) at least [average] score in the end-of-training satisfaction survey; (d) each participant delivers a [concrete output]. [Proportion] of the payment depends on meeting these acceptance criteria." This template is not only a document but also a negotiation tool. Reviewing the clauses of the template together with the supplier reveals early which side can be flexible on which point. A good supplier makes suggestions that strengthen your template; this is the first indicator of the supplier's expertise. ## AI Training Procurement in the Türkiye and KVKK Context When preparing an AI training technical specification, Türkiye's unique context brings both opportunity and responsibility. On the opportunity side is Türkiye's extraordinary speed of AI adoption; this raises organizations' appetite for building competency. On the responsibility side, the regulatory framework, KVKK above all, must be reflected in the specification. This high adoption in Türkiye is a double-edged sword for organizations. On one hand, employees already use AI tools personally; this offers ready ground for training. On the other hand, unmonitored personal use poses a risk to enterprise data security — employees may enter enterprise knowledge into external tools without realizing it. A well-designed training specification takes this dual reality into account: it both moves adoption into a formal and secure frame and teaches KVKK-compliant use. In the KVKK context, the specification must specifically address three scenarios. First, the processing of participant data: every training collects participants' personal data, and this falls under KVKK. Second, the use of enterprise data in training content: if the training has participants practice with the organization's own data, protecting this data is essential. Third, the KVKK compliance of the systems participants will build with what they learn in the training: the training should teach participants not only to use AI but to use KVKK-compliant AI. This third dimension adds a responsibility layer to the training's content and should appear as a learning outcome in the specification. In regulated sectors (banking, healthcare, telecommunications) this context deepens further. For example, BDDK regulations in banking, patient-data protection in healthcare, bring additional obligations. An organization in these sectors should also add sectoral compliance requirements to its specification and require familiarity with these regulations from the supplier. We cover the corporate framework of AI governance in what is AI governance and the principles of responsible use in what is responsible AI; these frameworks strengthen the compliance dimension of the training specification. ## How Does an AI Training Specification Differ by Role and Sector? The same AI training technical specification template is filled differently for different roles and sectors; because each role's and sector's need, priority, and learning outcomes differ. Adapting the template to context determines the training's accuracy. Below we address how it differs by both role and sector. ### For Senior Management and Decision-Makers For this audience, training is built on strategic awareness and decision-making competency rather than technical depth. Learning outcomes should be "evaluates the business value of an AI investment and prioritizes it," not "codes an AI model." Duration is kept short and intensive (executives' time is limited), and the format is usually in-person and interactive. In the specification, the presenting an AI project to senior management approach and, for the strategic framework, the how to build a corporate AI strategy guide can be referenced as complementary content. ### For Middle Managers Middle management is the critical layer translating strategy into practice; their trainings require a balance of both awareness and practical application. Learning outcomes should be both managerial and hands-on, like "designs and pilots an AI use case in their team's workflow." For this audience, use-case prioritization skill is an important outcome in the specification. To understand the corporate level of AI maturity, the corporate AI maturity model guide places middle management's training in the corporate context. ### For Technical Teams For technical teams (developers, data analysts), training requires depth and hands-on skill. Learning outcomes should be concrete technical competencies like "integrates a generative AI model into a workflow" or "designs and tests a prompt chain." For this audience prerequisites are higher (programming knowledge, tool access) and the curriculum goes into more technical topics. For basic technical concepts, guides like what is an LLM, what is a token, and what is RAG can be referenced in the specification as pre-preparation material for technical training. ### For All Employees (Broad Literacy) Some programs aim to give a broad, non-technical audience basic AI literacy. In this case learning outcomes should be simple and daily-work-focused: "uses AI safely and effectively in their daily work," "questions the reliability of an AI output." The format is usually scalable (online) and duration is kept short. For such broad programs, the what is AI literacy guide sets the training's frame. Sectoral differentiation also matters. In finance the emphasis on compliance and risk stands out; in manufacturing operational efficiency and predictive scenarios; in healthcare data sensitivity and ethics; in retail customer experience and personalization. The specification should reflect the sector's priorities in the learning outcomes and example scenarios. A generic "AI training," if it comes to every sector with the same examples, fits none exactly; sector-specific examples multiply the training's value. ## How Many Suppliers Should Receive the Specification and How Long Does the Process Take? A practical question is how many suppliers an AI training technical specification will be sent to after it is prepared, and how long the process will take. This directly affects procurement efficiency: too few suppliers weaken the comparison basis, too many make the evaluation load unmanageable. Experience shows that three to five suppliers are ideal for most corporate training procurements; this number both provides a meaningful comparison and keeps the evaluation reasonable. A pre-screening stage helps when determining the number of suppliers. First, potential suppliers are gathered from a broad list (market research, references, industry recommendations); then the obviously unsuitable ones (lack of experience, domain mismatch, insufficient scale) are eliminated to form a short list of three to five. The specification is sent only to this short list. This two-stage approach allows both choosing from a broad pool and keeping the evaluation focused. The duration of the process varies by the training's complexity but for a typical corporate AI training technical specification procurement can take from a few weeks to one or two months. This period covers the steps of preparing the specification, researching suppliers, collecting offers (suppliers should be given a reasonable time), evaluation, negotiation, and contract. Adding a question-answer window to the process — a period during which suppliers can ask questions about the specification — markedly improves offer quality; because suppliers who resolve uncertainties give more accurate offers. A rushed procurement process usually ends in a poor choice; that is why the schedule should be planned realistically from the start. After sending the specification, give suppliers the chance to ask written questions and share the answers to incoming questions with all suppliers. This transparency both ensures fair competition and reveals early the gaps in your specification. Often the questions suppliers ask are the most valuable feedback showing the weakest points of your specification. ## In-House or External Procurement? How Does It Affect the Specification? A fundamental question to answer before preparing an AI training technical specification is: should this training be procured externally, or delivered with in-house resources? This "make or buy" decision directly affects whether the specification is needed and how it will be shaped. Both paths have their logic, and the decision depends on the organization's maturity, scale, and strategic priorities. In-house training means the organization's own experts transferring knowledge. Its advantage is content fully aligned with the organization and a seemingly low cost. Its disadvantage is that internal experts are usually not good teachers, struggle to spare time from their own work, and cannot bring an outside perspective. Also, in-house training carries the risk of "organizational blindness": an organization cannot teach what it does not itself know. In a rapidly changing field like AI, the fresh view of external expertise is often valuable. External procurement is the path that requires an AI training technical specification. Its advantage is an expert instructor, current content, an outside perspective, and professional pedagogy. Its disadvantage is cost and the need to separately ensure alignment with the organization. For most organizations, the healthiest approach is hybrid: scaling basic and recurring trainings with in-house resources, and taking advanced, strategic, or fast-changing topics through external procurement. Even in this hybrid model, a sound specification is essential for the externally procured part. An interesting point: applying an AI training technical specification discipline produces value even if you will deliver the training with in-house resources. Writing learning outcomes, defining acceptance criteria, and planning assessment take in-house training too from amateurism to professionalism. The specification logic is a quality discipline independent of procurement. ## How Do You Move to Negotiation and Contract After the Specification? After the specification is sent and offers are evaluated with the scoring table, a frequently skipped but critical stage of the procurement process arrives: negotiation and contract. A good AI training technical specification has already laid the foundation of this stage; because every clause in the specification turns into a clause of the contract. However, negotiation is where the specification is turned into a contract and the final fine-tuning is done. There are three topics to focus on in negotiation. First, tying acceptance criteria and payment terms: how much of the payment depends on which acceptance criterion should be written clearly. Second, change management: what happens if scope or participant count changes during the training should be determined in advance. Third, liability and warranty clauses: if the training does not meet the acceptance criteria, which remedy mechanism (re-training, refund, correction) will kick in? When these three topics are clarified, the contract rests on a solid ground protecting both parties. For the negotiation to be healthy, one principle matters: negotiation is not cornering the supplier but building a common ground of success. A supplier pushed to the lowest price but whose motivation is broken usually produces the lowest value. A good negotiation looks after both the organization's budget and the supplier's sustainable profit margin; because a long-term, recurring training relationship is far more valuable than a one-off bargaining victory. The clarity of the specification also facilitates this negotiation: since the parties know what is what from the start, the negotiation turns from bargaining into fine-tuning. Even after the contract is signed, the specification's job is not done: as an annex to the contract, the specification remains a reference document throughout the implementation. When a dispute arises, the question "what did the specification say?" is the starting point of the solution. That is why the specification is a document written at the beginning of the procurement process but living to its end. To see, in a broader frame, why AI projects fail in general and the procurement/training dimension of that failure, the corporate AI maturity model and what is an AI roadmap guides tie training procurement to the holistic strategy. ## AI Training Specification Implementation Checklist The checklist below is a practical guide to preparing an AI training technical specification soundly from scratch. If you can check every item, your specification is ready to be sent to suppliers. Applying this checklist with a pilot training is smarter than trying to transform the whole organization at once. A small, well-defined pilot specification lets you both learn the process and mature the template before large-scale procurement. To build the AI roadmap at the corporate level, the what is an AI roadmap guide places training procurement into a broader transformation plan. ## What Are the Common Mistakes in an AI Training Specification? Seen with an experienced eye, most failed AI training technical specifications suffer from similar mistakes. The common feature of these mistakes is that they all increase uncertainty and make evaluation subjective. The most common are: - **Leaving scope vague:** Settling for "we want AI training" forces suppliers to guess and produces incomparable offers. Scope must be defined clearly together with audience and depth. - **Listing topics instead of outcomes:** A topic list like "machine learning, deep learning, LLMs" does not say what the participant will be able to do. The learning outcome is more important than the topic. - **Not defining the audience's level:** When the prior-knowledge level is not stated, the training feels either too basic or too complex and produces no value. - **Skipping the reference requirement:** Accepting an unproven competence claim is one of the most expensive procurement mistakes. A similar, current, verifiable reference should be required. - **Making price the only criterion:** Auto-selecting the lowest price often means selecting the lowest value. The scoring table balances quality with price. - **Leaving assessment and acceptance criteria unmeasurable:** A criterion like "the training must be successful" is useless. Acceptance criteria must be numerical and written in advance. - **Not adding KVKK and intellectual property clauses:** Leaving these clauses to after the contract creates both legal risk and knowledge leakage risk. - **Neglecting post-delivery reinforcement:** A program that ends when the training ends loses most of what was learned within a few weeks. Reinforcement and support should be in the specification. All these mistakes come down to the same root cause: delegating uncertainty to the supplier instead of resolving it. A vague specification pushes the supplier to choose either the cheapest or the easiest interpretation — which is almost never the organization's real need. A good specification reduces risk by reducing uncertainty. The most practical way to avoid these mistakes is to review the specification with an independent eye before sending it to suppliers. The value-add of a consultant experienced in AI training and transformation is exactly here: turning the organization's real need into correct outcomes, adapting the template to context, and objectifying the evaluation process. We cover why AI investments fail and its training dimension in reasons for failure in AI investments; good training procurement prevents an important part of these failure reasons from the start. ## What Are the Steps of the AI Training Procurement Process? An AI training technical specification sits at the center of a broader procurement process; you should think of the specification not alone but as part of this process. A sound procurement process consists of seven steps, from needs analysis to post-training evaluation, and each step builds on the previous one. **Step one — Needs analysis:** Procurement begins with a need. Which team has which competency gap? Is this gap closed by training, hiring, or a tool investment? A specification written without doing the needs analysis correctly tries to solve the wrong problem. In this step, the concrete business problem the training will solve is clarified. **Step two — Preparing the specification:** When the need is clear, the specification is written with the twelve components covered in this guide. The specification translates the need into a language suppliers can understand and bid on. The quality of this step determines the quality of the whole process. **Step three — Supplier research and pre-screening:** While the specification is being prepared, potential suppliers are researched and a short list is formed. Pre-screening eliminates obviously unsuitable suppliers from the start and makes the evaluation process manageable. **Step four — Collecting offers:** The specification is sent to the short-listed suppliers and offers are collected within a certain period. In this step, opening a question-answer window in which suppliers can ask questions improves offer quality. **Step five — Evaluation and scoring:** Offers are evaluated with the pre-fixed scoring table. In this step, having multiple evaluators score independently and then reconcile reduces subjectivity. **Step six — Negotiation and contract:** Details are negotiated with the highest-scoring supplier and the contract is signed. The clauses in the specification form the basis of the contract; the acceptance criteria and payment terms become binding here. **Step seven — Implementation and evaluation:** The training is carried out and its impact is measured with the assessment methods in the specification. This step closes the procurement loop and provides learning for the next training. Good organizations see training procurement not as a one-off purchase but as a continuously improving cycle. Each procurement leaves a learning that makes the next one better: which specification clauses worked, which supplier kept its word, which assessment method showed real impact? Documenting these learnings makes the organization increasingly mature in training procurement. Throughout the procurement process, the role of an AI consultant is to catch the points the organization might miss on its own: unrealistic learning outcomes, missing compliance clauses, or incomparable pricing structures. To structure your corporate AI training program correctly, you can review corporate training options, start with AI consulting for an organization-specific AI strategy and procurement framework, and deepen all basic concepts in the learning center. ## Frequently Asked Questions ### What is an AI training technical specification and why is it needed? An AI training technical specification is a formal procurement document (RFP) that defines, item by item, what an organization expects from suppliers when buying AI training. It is needed because the quality of an intangible service like training cannot be seen before the contract; the specification reduces this uncertainty. When you write down the scope, learning outcomes, instructor qualifications, assessment method, and acceptance criteria in advance, suppliers bid on the same basis and the later "this is not what we expected" dispute disappears. The specification turns a subjective purchase into an objective and defensible process. ### Which components must an AI training specification contain? A sound specification contains twelve components: (1) purpose and scope, (2) audience and prerequisites, (3) curriculum/content, (4) learning outcomes, (5) duration and format, (6) materials and platform, (7) assessment, (8) instructor qualifications, (9) reference requirements, (10) pricing structure, (11) delivery and logistics, (12) intellectual property, confidentiality, and KVKK. To these you should add measurable acceptance criteria and a supplier scoring table. If one of these components is missing, that gap turns into subjective interpretation during evaluation or a later dispute. ### How should learning outcomes be written in a specification? Learning outcomes should define what the participant will be able to do at the end of the training, with measurable verbs. Instead of unmeasurable phrases like "understands AI," write observable outcomes like "produces and prioritizes three concrete AI use cases for their own department." Bloom's taxonomy verbs (defines, applies, analyzes, evaluates, creates) guide this. The practical benefit of measurable outcomes is that assessment and acceptance criteria can be tied directly to them; so the question "was the training successful?" rests on evidence. ### How is a supplier evaluation scoring table built? A scoring table gives each supplier an objective total score by weighting evaluation criteria. Typically, technical competence (instructor experience, curriculum fit, references, methodology, material quality) and commercial offer (price, payment terms) are gathered into two main blocks and evaluated with, for example, 70/30 weighting. A score range (e.g., 1-5) and a weight are defined for each criterion; the supplier score is the sum of the products of criterion score and weight. This table prevents the lowest-price supplier from being auto-selected and gives a document-based answer to "why did we choose this supplier?" ### How should KVKK be handled in an AI training specification? The KVKK clause should cover the personal data to be processed during training (participant name, email, assessment results, sometimes internal sample data). The specification asks the supplier for these commitments: processing data only for training purposes, deletion or return after training, restrictions on transfer to subcontractors, data security measures, and a data processing agreement. If the organization's real data (e.g., actual customer records) will be used in training, an anonymization or synthetic data requirement should be added. Skipping the KVKK clause creates both legal risk and enterprise-knowledge leakage risk. ### How are instructor qualifications defined in the specification? Instructor qualifications should be defined with both formal (educational background, certifications) and evidence-based (real project experience, reference trainings, domain expertise) criteria. What is especially important in AI training is that the instructor has not only theoretical but production-environment AI implementation experience; because what participants value most is real-world examples. It is good practice to require the instructor's CV, if possible a teaching sample or short demo, and to state that instructor changes may be made only with the organization's approval. ### Should you ask for a fixed price or a per-participant price? Both have their place, and the specification should ask for both in a comparable way. A fixed package price (total fee for a given group size) provides budget predictability and is usually advantageous for small-to-medium groups. A per-participant price offers scalability when the number of participants is uncertain or very large. The specification should clearly list what the price covers (materials, platform, certification, transport, travel) and leave no hidden cost. Standardizing the pricing structure ensures offers are compared apples to apples. ### What are the most common mistakes in the AI training procurement process? The most common mistakes: leaving scope vague and settling for "we want AI training"; listing only topic headings instead of learning outcomes; not defining the audience's prior knowledge level; skipping the reference requirement; making price the single decisive criterion; leaving assessment and acceptance criteria unmeasurable; not adding KVKK and intellectual property clauses; and not putting post-delivery support/reinforcement in the specification. The common result of these mistakes is a training that is complete on paper but produces no real behavior change. ### How are acceptance criteria made measurable? Acceptance criteria are the measurable thresholds that must be met for the training to count as "completed and payable." Examples of good acceptance criteria: a minimum attendance rate (e.g., at least 80% of participants attending sessions), an assessment pass threshold (a given proportion of participants passing the assessment), a satisfaction score (a certain average in the end-of-training survey), and output delivery (each participant producing an implementation plan or project draft). When these thresholds are defined numerically in the specification, acceptance stops being a subjective "we liked it/we didn't" evaluation. ### Is there a ready AI training specification template? Yes, this guide provides a copy-ready example template item by item. The template contains fillable headings for each of the twelve components: purpose and scope, audience and prerequisites, curriculum, learning outcomes, duration and format, materials and platform, assessment, instructor qualifications, reference requirements, pricing structure, delivery and logistics, intellectual property/confidentiality/KVKK, and acceptance criteria. It is recommended to adapt the template to your organization's context (sector, team size, maturity level) and to replace general phrases with your own concrete needs. ## In Short: How to Write an AI Training Technical Specification? In short, the answer to how to write an AI training technical specification is: turn the organization's real need into twelve components that suppliers can bid on from the same basis and the organization can compare objectively. These components — purpose and scope, audience and prerequisites, curriculum, learning outcomes, duration and format, materials and platform, assessment, instructor qualifications, reference requirements, pricing structure, delivery/logistics, and intellectual property/confidentiality/KVKK — combined with measurable acceptance criteria and a supplier scoring table, turn a subjective purchase into a defensible decision. The most important message is this: a good AI training technical specification is not a document but a thinking discipline. Organizations that build that discipline manage training with evidence, not hope; they buy training for the right audience, with the right outcomes, and measurable results. For the basic concepts you can see the what is corporate AI training and what is AI literacy guides; for an organization-specific training program and procurement framework you can start with corporate training and AI consulting. A well-prepared specification turns your AI training from a formality completed on paper into an investment that creates real competency.