FDE Competency Framework v0.1

Turning a customer’s unclear, often contested request into a specific, valuable and feasible problem statement with agreed success measures, before significant build effort is committed.

1. 1.1Plans and runs discovery with users, operators, technical owners and decision-makers.
2. 1.2Maps the current workflow, systems of record and pain points as they really are, not as documented.
3. 1.3Separates the stated request from the underlying need, and identifies who actually decides.
4. 1.4Frames the problem as testable hypotheses, with baseline metrics and target outcomes.
5. 1.5Assesses feasibility and readiness (data, access, sponsorship, risk), and recommends go, reshape or stop.

Designing a solution that fits the customer’s real environment (legacy systems, identity, network limits, existing platforms) and choosing the simplest architecture that will deliver the outcome reliably.

1. 2.1Designs end-to-end architectures across the vendor platform, customer systems and third-party services.
2. 2.2Integrates with customer systems through APIs, events, files and databases, including legacy and undocumented interfaces.
3. 2.3Makes and documents build-vs-configure-vs-extend-the-product decisions and trade-offs.
4. 2.4Designs for the customer’s identity, access, network and data-residency constraints.
5. 2.5Chooses the right kind of solution, including deterministic code, classical ML, LLM or agent, or no software at all.

Building production-quality applications on AI models (including retrieval, tool use and agentic workflows) and proving their quality with evaluations the customer trusts.

1. 3.1Writes production-grade code (tested, reviewed, maintainable) in at least one mainstream language.
2. 3.2Builds LLM-based components: prompting and structured outputs, retrieval over customer content, tool and function calling, and multi-step or agent workflows.
3. 3.3Designs evaluations: test sets drawn from real cases, acceptance thresholds agreed with the customer, regression checks, and human review where needed.
4. 3.4Analyses errors and improves quality, cost and latency systematically.
5. 3.5Applies guardrails and keeps a human in the loop to match the risk of the use case.
6. 3.6Keeps up with changing model capabilities and judges when a change matters to a deployment.

Getting the customer’s real data (messy, siloed, governed, sometimes legacy) into a state where a solution can use it safely and reliably.

1. 4.1Profiles and assesses data quality, lineage and fitness for purpose.
2. 4.2Builds and operates pipelines (batch and streaming as needed) and data models.
3. 4.3Works within data governance: classification, consent, retention, residency, and least-privilege access.
4. 4.4Prepares unstructured content (documents, images, transcripts) for retrieval and extraction.
5. 4.5Plans migration and reconciliation from legacy sources, with checks the customer can verify.

Shipping solutions into the customer’s production environment and keeping them secure, observable, compliant and dependable, with security and compliance treated as part of delivery rather than an afterthought.

1. 5.1Packages and deploys software through repeatable pipelines (CI/CD and infrastructure as code) in customer-controlled environments, including restricted or air-gapped ones.
2. 5.2Instruments systems for monitoring, logging, alerting and AI-specific observability (quality drift, cost, latency).
3. 5.3Applies secure engineering: secrets handling, access control, vulnerability management, privacy, and threat modelling for AI systems (for example prompt injection and data exfiltration through tools).
4. 5.4Takes solutions through customer security, risk and compliance reviews.
5. 5.5Documents and runs responsible-AI controls appropriate to the use case: transparency, auditability, bias checks, and incident response.

Building trust across the customer organisation and making sure the solution is actually adopted, by explaining clearly, delivering bad news early, and managing the human side of change.

1. 6.1Adapts technical communication to executives, operators, engineers and risk owners.
2. 6.2Maintains relationships and trust, including when delivering bad news or pushing back.
3. 6.3Plans and supports adoption: training, workflow redesign, champions, and feedback loops.
4. 6.4Measures adoption and impact and reports them honestly.
5. 6.5Runs demos and readouts that tie technical progress to business outcomes, without overselling.

Running an engagement as a bounded, outcome-focused investment: scoping it, sequencing it, knowing when to stop, handing it over cleanly, and turning what was learned into reusable value for the vendor.

1. 7.1Scopes and sequences work to show value early (smallest proof first), and manages trade-offs between scope, speed and quality.
2. 7.2Defines and enforces entry, exit and stop criteria for an engagement.
3. 7.3Understands the commercial model of the engagement (cost to deliver, value delivered, intellectual-property boundaries, expansion potential) and acts accordingly.
4. 7.4Plans handover so the customer can run the solution without the FDE.
5. 7.5Codifies field learning into playbooks, reusable components and product feedback.