Forward-Deployed Engineering
Senior engineers,inside your organisation,until one workflow runs in production.
A named engineer joins your team with commit rights in week one, builds one workflow against your real records and hands it over as a running system. Six weeks, a fixed fee, your code.
- day 01 · access granted, repository cloned
- day 07 · first pull request merged
- day 14 · evaluation set agreed with operations
- Engagement
- One workflow, six weeks
- First pull request
- Day seven
- Code
- Yours from the first commit
- Fee
- Fixed before week one
FDE / 01The problem
Most pilots never leave the notebook.
The model was rarely the hard part. The hard part is the distance between a demonstration and a system a business can run on: access, connectors, evaluation, review, sign-off. That distance is where an engineer has to sit.
A pilot that worked in the demonstration
- Runs in a notebook on a sample export from last quarter
- Impressed the steering group; nobody has used it since
- No connector to the system of record
- No evaluation set, so nobody can say whether it is right
- The person who built it has moved to the next initiative
A system in production
- Code in your repository, reviewed by your reviewers
- Reads from the system of record and writes back to it
- Scored against an evaluation set before every release
- Monitored, with a runbook your team has executed
- An owner on your side who can change it without us
FDE / 02The model
Four things we hold to.
They are the same on every engagement, whatever the workflow. If one of them cannot be met, we say so before the first week, not after the sixth.
One outcome, scoped in writing
Before the engineer starts, the workflow, the metric that will judge it and the person who owns it are written on one page. Nothing else is in scope until that workflow is in production.
Commit rights in week one
The engineer works in your repository, under your access controls, from the first days. A merged pull request by day seven is how both sides know the access is real.
Built against an evaluation set, not a demonstration
Real cases with known answers are agreed with the people who do the work. Every candidate is scored against them before it goes near a user, and the score is the argument for shipping.
Production or it did not happen
The engagement ends when the workflow runs on real traffic with a signed runbook and a team that can operate it. A prototype, however good, is not a deliverable.
FDE / 03The offerings
11 ROWS
Eleven offerings, one way of working.
Each one puts an engineer inside your organisation with an outcome to reach. They differ in how long they run and in what you hold when they end.
- 01Readiness & Scoping SprintTwo weeks, one engineerA scoped workflow, a success metric and a fixed price
- 02Production SprintSix weeks, one named engineerA system in production, its evaluation suite and its runbook
- 03Embedded Engineering PodsQuarterly, two to three engineersA shipping cadence your team keeps after the pod leaves
- 04Pilot-to-ProductionEight to twelve weeksA pilot in production, with go-live criteria met and recorded
- 05Integration & Data OnboardingScoped per systemConnectors, a data model and retrieval with access walls intact
- 06Regulated & Secure DeploymentScoped per environmentA deployment with zero data egress and an audit trail
- 07Evaluation & ObservabilityRuns inside every engagementAn evaluation suite, monitoring and the thresholds that page someone
- 08Handover & EnablementThe last two weeks of any engagementA team that runs the system without us
- 09Managed OperationsMonthly, after handoverUptime, monitoring and iteration, reported monthly
- 10Engagement ModelsReferenceThe terms before the first call
- 11The EngineersReferenceWho turns up on day one
FDE / 04The engagement
W0 – W6
Six weeks, drawn as a plan.
Every workstream is a bar with a date at each end. The diamonds are the gates: scope signed, cut over, handed over.
Every bar ends in something you hold: a document, a merged branch, a passing suite, a signed runbook.
FDE / 05What you keep
6 ITEMS
Everything, in your accounts.
Handover is part of the engagement, not an add-on. These six items are checked off together before the engineer leaves.
- The repository
Every commit in your repository, intellectual property assigned before week one.
- The evaluation suite
The golden set, the scoring harness and the regression run, wired into your pipeline so a quality drop fails the build.
- The runbook
Deploy, rotate credentials, retrain, roll back, escalate; executed by your engineers before it is signed.
- The decision record
Each architectural decision, the options considered and why one was taken, dated.
- The accounts and keys
The system runs under your cloud account and your identity provider. We hold no standing access after handover.
- The trained owner
One named person on your side who has released, rolled back and retrained the system without us in the room.
FDE / 06The comparison
Against the other ways to get it done.
Each of these is the right call for something. The table says what each one gives you, including where the others are the better choice.
| Criterion | Forward-deployed | Consultancy | Staff augmentation | In-house hire |
|---|---|---|---|---|
| Ships production code | yes | no | partial | yes |
| Works inside your repository | yes | no | yes | yes |
| Fixed scope and fee | yes | partial | no | no |
| Owns the outcome | yes | no | no | yes |
| Leaves a team that can run it | yes | no | partial | yes |
| Starts within weeks | yes | yes | yes | no |
| Knows your systems on day one | no | no | no | yes |
FDE / 07Fit
Whether to call us
Call us when
- One workflow matters and has a date attached to it
- You can grant repository and system access inside a fortnight
- Someone senior owns the outcome and can clear a blocker in a day
- A pilot exists and has stalled between demonstration and production
- You would rather have one workflow in production than six in pilot
Do not call us when
- You need a strategy paper or a vendor evaluation first
- Your team needs hands on a backlog, not an outcome
- The workflow has no owner on your side
- No one can grant repository access in the first weeks
- Nobody internally has the capacity to take the handover
FDE / 08Who turns up
10 FIELDS
The engineers
The same rules for every engineer we place, written down so you can hold us to them.
- Seniority
- Senior only; each has taken systems into production and run them afterwards
- Named
- You meet the engineer during scoping and receive a written profile; the same person stays through handover
- Availability
- One engagement at a time, full time on your work; a start date given before you sign
- Where they sit
- In your repository, your standups and your review queue; on site where the work needs it
- Reporting
- To the owner of the outcome on your side, weekly, in writing
- Tools
- Yours: your pipeline, your reviewers, your access controls, your ticketing
- Languages
- Whatever your stack runs; we bring no platform to license
- Access
- Under your identity provider, scoped to the work, revoked at the handover gate
- The backup
- A second named engineer who has read the scope and the decision record, and who you meet if they are ever needed
- After the engagement
- Second on your on-call rota for an agreed period, then deliberately less involved
FDE / 09What you know first
6 ITEMS
Six things, before a signature.
The profile you receive after the scoping call. Nothing is signed until every item on it has a name or a date against it.
- The name
The engineer who will do the work, met during scoping. Not a role, not a team, not a bench.
- Prior production systems
The systems the engineer has taken into production and run afterwards, described in enough detail to ask about them.
- Availability
The date the engineer is free to start, and confirmation that nothing else is scheduled for the length of the engagement.
- The reporting line
Who the engineer reports to on your side, how often and in what form; the owner of the outcome, weekly, in writing.
- The backup
The second named engineer who has read the scope and would step in, and the condition under which you would meet them.
- The exit
The date of the handover gate, what passes it, and when the engineer's access ends.
FDE / 10The shapes
Four shapes, compared.
The same criteria down the side for each shape. Where a cell is partial or no, that is the shape, not a gap: a scoping sprint does not ship code and operations does not end at a gate.
| Criterion | Readiness sprint | Production sprint | Embedded pod | Managed operations |
|---|---|---|---|---|
| Fixed fee, agreed before it starts | yes | yes | yes | yes |
| Fixed duration | yes | yes | partial | no |
| Named engineer, met before you sign | yes | yes | yes | partial |
| Ships production code | no | yes | yes | partial |
| Ends in a handover | partial | yes | yes | partial |
| Recurring | no | no | yes | yes |
| Starts within weeks of the call | yes | yes | yes | no |
FDE / 11What we do not do
Four things, refused.
Each of these is common in the market. Each is a sign that the accountability is for time spent rather than for an outcome, so we do not offer it, even when asked.
Hourly billing
An hourly rate pays for approaching the outcome, not for reaching it, and it rewards the slower path. Every shape has a fixed fee agreed before week one against a written end. If the work takes longer for reasons on our side, the fee does not move.
Open-ended retainers
A retainer with no scope is a bank of hours by another name. Operations has a written scope, a monthly report and a notice period; a pod has a quarter and a gate. Nothing runs on without a written end or a written renewal.
Invoicing an unmet milestone
An invoice is raised when a gate is passed, not when a date is reached. If the gate is missed for reasons on our side, we stay until it is met at no further fee. If the cause is on your side, the clock pauses and we write down where it stands.
Subcontracting
The engineer you met at scoping is the engineer who turns up. We do not place a different person, and we do not pass the work to a third party. If the engineer cannot continue, you meet the replacement before they start, and you can decline.
FDE / 12Specified
10 FIELDS
The terms, specified.
As they stand before the first call. Scoping fixes the detail for one engagement; it does not change these.
- Readiness & Scoping Sprint
- Two weeks, one engineer; ends in a scoped workflow, a metric and a fixed price for the sprint that follows
- Production Sprint
- Six weeks, one named senior engineer; ends at the handover gate with a system in production
- Embedded Engineering Pods
- A quarter at a time, two to three engineers in your repository; renewed in writing, each quarter ending in a handover
- Managed Operations
- Monthly after handover, a flat fee; ends any month with notice
- Fee
- Fixed for every shape, agreed before week one; the scoping sprint prices the production sprint
- Invoicing
- At the gate, not the date; a missed gate on our side is not invoiced
- Scope
- One page: the workflow, the metric, the owner, the exit gate; changes are written down for the next engagement
- Intellectual property
- Assigned to you before week one; no platform to license, no runtime to rent
- Ending early
- A sprint can be stopped at any gate with what exists so far handed over; operations ends with notice from any month
- Starting
- A scoping call with an engineer; a written scope; then the fee and the date
FDE / 13Proof
16 ROWS
Case studies
Illustrative engagements written end to end, chapter by chapter.
- 01RestaurantsDemand forecasting, prep sheets and supplier ordersPre-consumer waste 7.9% → 6.1% of food cost; stock-outs halved
- 02RestaurantsPhone ordering by voice, in the caller's language49% of order calls completed with no person; peak calls answered in 20s 68% → 95%
- 03RetailSupplier catalogue onboarding and enrichmentSubmission to live listing 16 → 5 days; no legal field ever written by a model
- 04ManufacturingSupplier invoices, inbox to posted ledger entry61% of invoices posted with no keying
- 05ManufacturingVisual inspection of machined castingsCustomer-reported defects 41 → 11 ppm; false holds 3.6% → 1.3%
- 06ManufacturingMaintenance: alarms and notes to planned workAlarm to approved work order 19h → 3.1h; safety text never paraphrased
- 07ManufacturingWeekly production scheduling with a solverPlanner time 14h → 3.5h; on-time-in-full 72% → 84%
- 08Shared servicesIT and HR helpdesk agent with bounded actions38% of contacts resolved with no person; wait 4.1h → under 2 min
- 09DistributionCustomer purchase orders into ERP sales orders54% of email orders released with no re-keying; confirmation 6.4h → 1.2h
- 10StaffingContractor timesheets and compliance filesPayroll cycle close 6.0 → 2.5 days; no batch paid without a named approver
- 11LogisticsShipment exception handling at a 3PLTime to customer update 4.1h → 0.9h; claims filed in window 71% → 96%
- 12Real estateCommercial lease abstraction and obligationsAnalyst time per lease 4.5h → 1.3h; every term cites clause and page
- 13HealthcareCashless pre-authorisation packetsEmergency pre-auth submitted in 55 min, down from 2.4h
- 14InsuranceMotor claims first notice of loss and triageComplete file with a handler 3.8 → 1.1 days; no claim denied by the system
- 15LendingPeriodic KYC re-verificationOverdue KYC population roughly halved; officer time per file 26 → 7 min
- 16EducationAdmissions document processingSubmission to verified 9 → 1.5 days; no applicant rejected or ranked by the system
FDE / 14Questions
18 QUESTIONS
Asked before we start.
What is a forward-deployed engineer?
A senior software engineer who works inside your organisation, in your repository and your standups, to take one named workflow into production. The accountability is for an outcome that ships, not for hours spent approaching it.
How is this different from a consultant?
A consultant delivers a recommendation and leaves you to build it. A forward-deployed engineer builds it, in your codebase, and stays through cut-over and handover. If what you need is a strategy or a vendor evaluation, a consultancy is the right call and we will say so.
Do the engineers work on site?
Where the work needs it, yes. Scoping and the first days of embedding are usually on site, because access and the people who do the work are there. The rest runs wherever your team runs, in your tools and your cadence.
What happens to the code?
It is yours from the first commit. It lives in your repository, intellectual property is assigned before week one, and there is no platform to license or runtime to rent. If you never speak to us again, nothing stops working.
How is the workflow chosen?
Together, during scoping, with the person who owns the outcome and the people who do the work. We look for one workflow that matters, has a measurable result and can be given real access quickly. The choice is written on one page before the engineer starts.
What if it is not ready in six weeks?
The exit gate is written down at scoping, so both sides can see it coming. If the gate is missed for reasons on our side, we stay until it is met at no further fee. If access or a decision on your side is the cause, we stop, write down where it stands and agree what it would take to finish.
How are security and data handled?
The work runs inside your perimeter: your cloud account, your identity provider, your region, or on your premises where required. Nothing is copied out to train anything. Every action the system takes is attributed and logged, and we hold no standing access after handover.
How do we start?
Book a scoping call. It is thirty minutes with an engineer, not a salesperson, and you leave with a view on whether the workflow fits and what it would take. If it does not fit, you hear that on the call.
Can we interview the engineer before we sign?
Yes, and you should. You meet the engineer during scoping, usually in the Readiness & Scoping Sprint or on the scoping call, and you receive a written profile. If it is not the right fit, you say so before the scope is signed and you meet another.
What does senior mean here?
That the engineer has taken systems into production and run them afterwards: carried the pager, rolled back, retrained, written the runbook. It is a description of what they have done, not of years, and the profile lists the systems so you can ask about them.
What if the engineer leaves part-way?
The backup named on the profile has read the scope and the decision record from the start. You meet them before they begin, and you can decline. The gate, the fee and the scope do not change.
Are they engineers or consultants?
Engineers. They write code in your repository, ask your reviewers for approval and carry the consequences of what they ship through cut-over. If what you need is a recommendation rather than a system, a consultancy is the right call and we will say so.
How many engineers are on an engagement?
One, for a sprint; it is the shape of the work, and one person who knows the whole system is what makes the handover clean. Embedded Engineering Pods are two to three engineers, and each of them is held to the same rules.
Where are the prices?
On the scope, once there is one. Every fee is fixed and agreed before week one, and the scoping sprint exists to arrive at that number for the production sprint. A price quoted before the workflow is known would be a guess, and you would pay for the guess.
Which shape do we start with?
If the workflow is chosen and access can be granted quickly, the Production Sprint. If the workflow is not chosen, or access is uncertain, the Readiness & Scoping Sprint first; it is two weeks and it ends with the sprint scoped and priced.
Why is there no hourly option?
Because an hourly rate makes the accountability about time rather than the outcome, and it is the tell of a staffing arrangement rather than an engineering one. If what you need is hours on a backlog, a staffing firm is the right call and we will say so.
Can we stop part-way?
At any gate, yes. What exists at that point is handed over as it stands: the code in your repository, the scope, the decision record, the golden set so far. Operations ends with notice from any month, with the handover manifest re-run.
Do you work through partners or subcontractors?
No. The engineer you meet at scoping does the work. If they cannot continue, you meet the replacement before they start and can decline. Nothing is passed to a third party.
FDE / ENDStart
Start with one workflow.Judge us by where it ends up.
Thirty minutes with an engineer is enough to say whether the workflow fits and what it would take.