Business discovery
Find the right business problem
Understand goals, processes, data and constraints across organisations and departments.
Practise workflow interviews, problem framing and prioritisation.

FORWARD DEPLOYED ENGINEER
Understand a business. Own a project. Deliver an AI system that runs.
For people with technical or enterprise project experience. Work with a mentor on business discovery, requirements and agentic AI, building work you can demonstrate, evaluate and hand over.
5 places each month · Enquire about this month
Mentors and times subject to availability

One to one, on your project
BUSINESS × ENGINEERING × DELIVERY
Forward Deployed Engineers work closely with customers, translate ambiguous problems into technical solutions and help systems get adopted. Business, engineering and collaboration come together.
Business discovery
Understand goals, processes, data and constraints across organisations and departments.
Practise workflow interviews, problem framing and prioritisation.
Stakeholder management
Clarify expectations with business, engineering and decision makers. Challenge requirements and negotiate scope.
Practise discovery meetings, proposals, change and trade-offs.
Hands-on AI engineering
Use Python, LLMs, tools and APIs to build a reliable, controlled workflow.
Practise architecture, implementation, evaluation and approval.
Ownership & delivery
Clarify uncertainty, remove blockers and balance speed, quality and adoption.
Practise milestones, risk communication, acceptance and handover.
WHO THIS IS FOR
Discuss your background and goals before starting. Scope and pace are adapted to your skills and available time.
Your starting pointYou already code, develop features or build prototypes.
Mentoring focusAdd business judgement, customer communication and delivery.
Your starting pointYou have customer, solution design or integration experience.
Mentoring focusDevelop hands-on agentic AI implementation and evaluation.
Your starting pointYou know business problems and are willing to build.
Mentoring focusPractise a structured path from ambiguity to working systems.
YOUR 12-WEEK JOURNEY
This is the standard progression. Agree personal goals, scope and acceptance criteria at the start, then adapt to real progress.
01Discovery / Stakeholders / Scope / Judgement
02Python / LLMs / Agentic AI / APIs
03Acceptance / Handover / Case study / Interview
Scenario illustration
MONTH ONE
Month-end goal: explain the business problem, stakeholders and boundaries, with an actionable plan.
Week 01
Week 02
Week 03
Week 04
MONTH TWO
Month-end goal: a runnable core flow with documented decisions and evaluation.
Week 05
Week 06
Week 07
Week 08
MONTH THREE
Month-end goal: work you can open, evidence you can inspect and an accurate project narrative.
Week 09
Week 10
Week 11
Week 12
HOW WE WORK TOGETHER
Mentors guide and review. You own implementation and follow-through. Agree what to deliver, who can help and how to address blockers.
Bring current work, decisions and questions to make each session specific.
Review requirements, architecture or code; practise customer conversations and prioritise.
Complete agreed work, submit it for feedback and track unresolved issues.
Normally weekly; dates confirmed together.
Review agreed documents, architecture, code or delivery material.
Generally respond within two working days; complex topics move into sessions.
Targeted resume revision and mock interviews / project defences within the 12 sessions.
YOUR MENTOR TEAM
Each session pairs one mentor with one learner. Confirm the first phase before starting; any change includes progress, decision and feedback handover.



A named mentor is not guaranteed throughout. All three do not attend every session together.
WHAT YOU BUILD & KEEP
Work towards the agreed scope and compile a demonstrable, inspectable delivery pack. Record unfinished work and limitations accurately.
Open and operate the agreed core flow in a demo or local environment.
Problem, stakeholders, priorities, scope, acceptance and change.
Architecture, interfaces, configuration and implementation trade-offs.
Results, failure cases, cost boundaries, limitations and operating steps.
Context, responsibility, solution and outcomes, redacted where necessary.
Resume, project narrative, defence feedback and development plan.
OPEN THE EXAMPLE
Outcomes for case company Harbour Supply, showing the content and depth of the project deliverables.
Read the complete problem, stakeholders, trade-offs, architecture, acceptance and handover.
Case company Harbour Supply receives delivery, damage and refund requests. This system classifies requests, references policy, drafts replies and queues approval. It does not send email, issue refunds or change orders.
Support owns response quality, the warehouse checks shipment and damage facts, and finance controls refunds. A request to close all tickets automatically is narrowed to draft creation and human review because model output cannot establish logistics facts or refund authority.
| Requirement | Allowed system behaviour | Boundary |
|---|---|---|
| RQ-01 Delivery enquiry | Classify and draft a tracking enquiry response | Do not infer arrival dates without carrier records |
| RQ-02 Damage report | Request photos and route to support and warehouse | Do not assign fault or promise replacement |
| RQ-03 Refund request | Collect order, reason and item condition | An authorised person makes the refund decision |
| RQ-04 Reply approval | Present evidence and draft for review | Approval and sending are separate; nothing is sent here |
A Python API validates fields and access. An LLM proposes a category and retrieves versioned policy. A structured validator checks evidence and required fields. Missing evidence and financial decisions go to human review. Logs record sources, category, approver and state.
The model proposes; deterministic validation checks structure and evidence; authorised people decide. Missing evidence or retrieval failure routes to a person rather than continuing automatically.
The three deterministic cases below include category, policy, next action and draft. Damage and refunds remain under human control. None are sent. This is a structural check, not measured model accuracy, latency or business benefit.
| Input | Delivery requirement | Existing outcome |
|---|---|---|
| HS-1042 Delivery enquiry | No delivery promise without tracking evidence | Category, D-01 reference and a verification draft |
| HS-1043 Damage report | Collect evidence before confirming a remedy | Category, D-02 reference and photo request draft |
| HS-1044 Refund request | An authorised person must review the refund | Category, R-01 reference and information request draft |
These items inspect the completeness of authored examples. Real model accuracy, latency and business improvements have not been measured; no production performance is claimed.
Token usage, throughput and cloud costs have not been measured. A real integration needs least privilege, redaction, vendor approval, timeouts and fallback.
Handover covers API contracts, policy versions, evaluation cases, configuration and review procedures. Before production, the business must confirm policies and access, evaluate authorised representative data and test failure and rollback paths. This is not production acceptance evidence.
BEFORE YOU START
It is primarily for people with technical or enterprise delivery experience. Basic programming, APIs, Git and English technical reading are recommended. Discuss suitability first; prerequisite study may be recommended.
Startup experience is helpful, not required. Use authorised business materials where available, otherwise clearly labelled simulations. Real clients, enterprise access and production deployment are not included by default.
Allow around 8–12 hours of independent project work beyond the weekly 90-minute session. Effort varies with skills and complexity. You own the implementation; mentors guide and review.
Lightman, Joe and Jason are scheduled according to availability and project needs. Confirm arrangements before starting. A particular mentor is not guaranteed throughout, and all three do not attend together.
Agree a feasible scope and milestones at the start. Completion depends on skills, effort, complexity, data and access. Record finished work, limitations and outstanding tasks honestly.
Any third-party tools, model APIs and cloud resources are discussed before use. Joining does not automatically enable paid resources.
No employment, interview, offer or salary is guaranteed. The service provides the agreed guidance, review and support. Hiring outcomes depend on individual circumstances and the market.
Yes. Discuss your background, goals, project and availability before agreeing suitability, mentor schedules, service terms and start date.
LET’S TALK
Limited to 5 places each month. Contact us to confirm current availability and discuss your goals, project direction and mentor schedule.