FDE Council Insights · Standards and certification

How do you map a learning path to the FDE framework?

Map each part of a learning path to one of the 11 core domains in draft v0.3. Add AI and agent engineering only when the work uses models or agents. Keep a tool stack, hiring exercises, and clearances outside the framework, and call the path aligned to the draft.

By FDE Council · Published · Updated

In brief: The 11 domains are the backbone of a learning path, and they are not enough on their own. Draft v0.3 also names AI and agent engineering for deployments that use models or agents, and it still has no level-by-level indicators for the 11 domains. A path may say it is aligned to that draft. It may not say anyone is certified against it.

A learning path can be aligned to draft v0.3. It cannot be certified against it.

What does a path say on the first page?

It names the draft, the level it aims at, and the limit on what that draft can support.

Use this wording, or a shorter form of it: “Aligned to the FDE Council Professional Capability Framework v0.3 (draft).” Pair it with the status of the draft. Draft v0.3 has not been validated by a job task analysis. It is not a certification standard, and nobody can be assessed or certified against it. There is no Forward Deployed Engineer certification yet. The free self-assessment scores the 11 domains of this draft. It does not score AI and agent engineering as its own domain.

Say which level the path aims at. Practitioner is the right default: one deployment, owned from end to end. Associate is scoped work under guidance. Lead is several deployments, or one high-stakes deployment, plus setting the approach for others. Years of experience do not define the levels. Trust with customer executives describes Lead scope. The skill of adapting communication for executives sits in Domain 9.

Earlier drafts stay available and were not rewritten. v0.2 keeps the October 11-domain text. v0.1 is the seven-domain text that still holds the only published behavioral indicators. Cite v0.1 when you use those indicators. Do not present them as v0.3.

Which of the 11 domains does the path cover?

All eleven. Tag every core module with its domain number, including the domains job postings rarely name.

A path that copies only what postings mention will be strong on discovery, architecture, software, collaboration, and product feedback, and thin on workflow design, data, evaluation of non-AI systems, production operations, security, and commercial judgment. Those last six are still core. The skills article already shows that postings and the framework are not the same list.

What a learning path covers in each core domain of draft v0.3, and where level detail already exists
DomainWhat the path coversWhere level detail exists
1. Discovery and problem framing Who decides, and a go, reshape, or stop recommendation when data, access, sponsorship, or feasibility is missing v0.1 domain 1
2. Workflow design and value definition The future workflow, human decisions, exceptions, a measurable benefit, and the conditions for success Self-assessment only. v0.3 has no indicators
3. Solution architecture and enterprise integration The choice to build, configure, or extend the product, including when no new software is the right answer v0.1 domain 2
4. Data foundations and engineering What the data means, who may use it, quality, and dependable models, transformations, and pipelines v0.1 domain 4
5. Software engineering and rapid experimentation Reading unfamiliar code, building a working solution, and testing a risky assumption quickly Self-assessment only
6. Verification, evaluation and outcome measurement Whether it works, fails safely, is used, and improves the work, including for a system that does not use AI. When it uses a model, add a test set, allowance for answers that vary, and a regression check after the model changes Self-assessment, plus the measurement article. v0.1 domain 3 covered AI evaluation only
7. Production deployment, reliability and operations Safe deploy, monitoring, recovery, operating cost, including token and inference cost when a model is called, a runbook that names the model version, knowledge transfer, support, and a named owner Part of v0.1 domain 5. Handover also appears in v0.1 domain 7
8. Security, privacy, safety and governance Risks, controls, approvals, transparency, a record of how a decision was made, and human review where a person must intervene Part of v0.1 domain 5
9. Embedded collaboration, trust and adoption Communication with executives, operators, and engineers, disagreement, and changes to roles and workflows v0.1 domain 6
10. Delivery and commercial judgment Scope, trade-offs, not overpromising, when to narrow or stop, and judging cost against value v0.1 domain 7, except expansion, which v0.3 leaves out of the core
11. Field learning, product feedback and reuse Field lessons turned into product feedback and reuse, within confidentiality and product boundaries v0.1 domain 7, competency 7.5. No separate indicator table for this domain

Read v0.1 indicators as the older draft’s behaviors, not as a silent update to v0.3. v0.1 domain 7 includes expansion potential. Draft v0.3 says expansion sits with account management, which is a different role. Leave expansion out of a path that claims to follow v0.3.

The self-assessment asks two questions in each of the 11 domains and offers practice tasks. Use it. It scores draft v0.3, including a go, reshape, or stop call and the choice to build, configure, or extend. It does not score AI and agent engineering as a twelfth domain. For Domain 6, use the four rows in the measurement article on at least one system that does not use AI: works, fails safely, is used, and improves the work. When the system uses a model, also use an agreed test set, allow for answers that vary, and recheck after the model changes.

Where do language models and agents go?

They go in AI and agent engineering, a context-required capability. They do not become a twelfth core domain.

Use that capability when the deployment uses a language model, retrieval over customer content, tool calling, or a multi-step agent. Skip it when the deployment does not. AI and agent engineering in draft v0.3 asks for prompting and structured outputs, retrieval, tool calling, a person who can intervene before the agent acts on a customer system or a customer decision, mitigation of prompt injection, and control of customer data leaving through a tool. It also names the model version, version drift, and a replacement if the provider retires the model. Prompt injection cannot be fully prevented. Domain 5 still covers the surrounding software. Domain 6 covers whether it works, and for a model that includes a test set, allowance for answers that vary, and a regression check after the model changes. Token and inference cost stays in Domain 7. Domain 8 still covers security, privacy, and approvals.

Postings ask for this work often, and they use narrower labels than the framework. Bloomberry, in an analysis of 1,000 Forward Deployed Engineer postings supplied by Revealera, reports AI agents in 35% of postings, LLM experience in 31%, and RAG in 12% (Bloomberry, updated January 25, 2026; sample window not stated). The framework’s word for RAG is retrieval. A path may say RAG. It should also say retrieval, so the label matches the draft. Lightcast’s September 15, 2026 article names artificial intelligence, Python, and software engineering among the skills employers ask for, and it does not print chart percentages in the article text (Lightcast). The chart shares, including artificial intelligence at 70% and Python at 57%, are the reading in the Council’s skills article from October 30, 2026.

Where do tools, hiring, demos, and industry knowledge go?

Outside the 11 domains. Some of them sit next to a domain. None of them is a new domain, and none of them is a certification.

Topics a learning path often adds, and where draft v0.3 places each one
TopicWhere it goesWhat it must not say
Language models, retrieval, and agents AI and agent engineering, when the deployment uses them That this is a twelfth core domain, or that every FDE must take it
A named stack The path’s own practice vehicle. Label it as the path’s choice That the framework names Python, TypeScript, SQL, or a cloud
A demo or a workshop Domain 5 when it tests a risky assumption. Domain 6 when it is evidence. Domain 9 when it supports a change in how people work That a demo is a successful deployment, or that winning the next contract is Domain 10
Hiring exercises Outside the framework. Use the hiring article That a hiring exercise is a specialization, a credential, or a Council assessment
Clearances and sector knowledge Outside the draft. Use air-gapped environments and regulated-data constraints when those limits apply That the draft has an industry track, or that a clearance is one of those two rows

The framework names no tools and no vendors. That is deliberate. A path may still pick one stack so people can practice Domains 4, 5, and 7 on something concrete. Bloomberry’s page text, accessed November 7, 2026, reports Python in 66% of the 1,000 postings, TypeScript in 35%, AWS in 32%, GCP in 22%, and Azure in 18%. The Council’s October 30 skills article recorded the same analysis at one decimal place, including Python at 66.2% and TypeScript at 35.2%. Use either reading, name the source, and label the stack as the path’s choice. SQL appears in that skills article’s domain map. This article does not restate an SQL share, because the Bloomberry page text opened on November 7 does not print one.

A demo counts in the path when it tests an assumption or shows the agreed outcome. It does not count as success. The measurement article says a successful deployment is a system in production that a named customer owner can run, with an outcome that moved against a baseline. Turning a pilot into the next contract is account management. Draft v0.3 leaves case studies, account growth, and expansion out of the core. Bloomberry notes that some postings emphasize demos and sales support, and that the work it treats as Forward Deployed Engineering is production engineering rather than pre-sales.

Hiring is a different task from doing the job. The hiring article says what an employer can ask for, and it says the framework is not a hiring test. System design, a coding exercise, a customer role-play, and a delivery-judgment case can support Domains 3, 5, 1, 9, and 10. They stay outside the framework. Do not call them the specialization layer. Draft v0.3 names two examples: portfolio practice leadership, which builds on Domains 9, 10, and 11 at Lead scope, and AI evaluation lead, which builds on Domain 6 and AI and agent engineering. Neither is required of every FDE.

Clearances and sector primers are not in the draft. v0.3 says industry-specific tracks are not in this draft. Bloomberry’s industry section includes defense and healthcare examples. Those examples show that sector context shows up in postings. They do not add a track to the framework. When a deployment cannot assume open network access, use the air-gapped row. When the data brings residency, consent, retention, or sector rules beyond ordinary access control, use the regulated-data row. A clearance is neither row.

Which level should the path aim at?

Practitioner, unless the path says it is for Associate or Lead and shows the evidence for that scope.

v0.3 describes the three levels as scope, not as a syllabus. Associate delivers a defined part under guidance. Practitioner owns one deployment. Lead owns several, or a high-stakes one, and sets methods other people reuse. The draft does not say, domain by domain, what to teach or how to test each level. Until it does, a path that needs that detail uses two labeled stand-ins. For domains 1, 3, 4, 9, and 10, the matching v0.1 indicator table is the stand-in, and the path says “v0.1 indicators.” For domains 2, 5, 6, and 11, and for the parts of domains 7 and 8 that v0.1 bundled into one domain, the self-assessment anchors are the stand-in, and the path says “v0.3 self-assessment.” Those anchors are a check of the 11 domains, not a full indicator table.

Use this checklist before calling a path aligned to the draft:

  1. Name draft v0.3 and link the framework.
  2. Say “aligned to the draft.” Do not say certified, accredited, approved, or official.
  3. Put a domain number on every core module, including Domains 2, 4, 6, 7, 8, and 10.
  4. Add AI and agent engineering only when the work uses models or agents, and keep the test of whether it works in Domain 6, including one system that does not use AI.
  5. Label any stack as the path’s choice. Do not attribute the stack to FDE Council.
  6. Keep hiring exercises, a contract win, clearances, and sector primers outside the framework, and point to the page that covers each allowed neighbor.
  7. If you use v0.1 indicators or the self-assessment, name that version in the module.

Frequently asked questions

Can a learning path say it is certified against the FDE framework?

No. Draft v0.3 has not been validated by a job task analysis. A path may say it is aligned to FDE Council Professional Capability Framework v0.3 (draft). It may not say certified, accredited, approved, or official. There is no FDE certification yet.

Should AI and agents be a twelfth domain?

No. AI and agent engineering is a context-required capability. Use it when the deployment uses a language model, retrieval, tool calling, or a multi-step agent. Skip it when the work does not. Whether the system works stays in Domain 6.

Does the FDE framework name Python or a cloud?

No. The framework names no tools and no vendors. A learning path may pick one stack as its practice vehicle and must label that stack as its own choice. Bloomberry reports Python in 66% of 1,000 postings and TypeScript in 35%.

Where do hiring exercises belong?

Outside the framework. Use the Council’s hiring article for what an employer can ask. Do not call a hiring exercise a specialization or a certification. Draft v0.3 names two examples: portfolio practice leadership, and AI evaluation lead. Neither is required of every FDE.

Which framework version should a learning path cite?

Cite draft v0.3 for the current 11 domains and the context-required capabilities. Cite v0.1 only when you use its behavioral indicators, and say so. The free self-assessment scores the 11 domains of draft v0.3. It does not score AI and agent engineering as a separate domain.

Sources

How to cite this article

FDE Council. (2026, November 7). How do you map a learning path to the FDE framework? https://www.fdecouncil.com/insights/map-learning-path-fde-framework/. Please also cite the original source of any figure you reuse.

Basis: FDE Council draft framework v0.3 (updated November 7, 2026), archived drafts v0.1 and v0.2, the self-assessment for the 11 domains of draft v0.3, and the Bloomberry and Lightcast pages accessed November 7, 2026. The domain map is a reading of those drafts, not a validated standard. Bloomberry’s rounded page text and the October 30 one-decimal reading are both reported. The Council’s text and tables are licensed CC BY 4.0; third-party figures remain under their sources’ terms. Not employment or financial advice.

More FDE Council Insights · Read the open competency framework

Related: What FDE job postings ask for · How employers can assess FDE skills · How to measure whether an FDE deployment succeeded · What the FDE delivery lifecycle looks like.