At portfolio scale
Ten prototypes narrow to one production decision
Our other engagements cover one workflow at a time. An institution with twenty candidate workflows runs the same method more often and in parallel, with a written decision at every seam, and needs no multi-year program. This page describes that shape and what it commits you to.
The method does not change
One loop, run more often
A portfolio program repeats the same four steps and ranks the workflows once, at the front. The method inside is the one a single engagement uses.
Four steps run in sequence. Scope, what it will attempt and what it will not. Build, on your material in an environment you name. Test, real cases with known answers and the misses counted. Decide, in writing, with the option to stop at no further cost. The sequence then repeats for the next workflow. A program of ten prototypes and five pilots runs this same loop fifteen times.
Scope, build, test, decide. That sequence is what a governed prototype does in four weeks and what a knowledge agent pilot does in six. Running fifteen of them adds no new method, contract shape or governance model. It adds a queue.
The AI roadmap produces the queue. It runs four weeks and ends in a ranked sequence of candidate workflows, with the value case, the control requirements and the cost of each one written down. The ranking has to happen before anything else, and it costs least.
Does anything change at portfolio scale?
Three things change. Prototypes run five at a time, so peak staffing is higher. Selection happens across the whole set, so a prototype that finishes early waits for its cohort. We then compare the decisions against each other, which is the reason to run ten instead of one.
The shape
Two rounds of selection, each one on test results
A prototype is a four-week build answering whether a workflow can work at all. A pilot is a six-week build on your real cases answering whether you should run it. Both end in writing, so stopping is a result.
Show the detail
One roadmap runs for four weeks and produces a ranked sequence. Ten governed prototypes follow, four weeks each and five at a time; five carry on and five stop. Five governed pilots follow, six weeks each with weekly checks; one carries on and four stop. The buyer approves one system for production. We price the production release separately and then operate the system monthly.
Ten prototypes produce ten written recommendations. Five of those recommend stopping, and the other five continue on evidence from a build. Five pilots produce five acceptance reports. Four recommend stopping. One reaches a production decision.
Nine of the fifteen builds therefore end in a documented decision not to proceed. That is what the shape is for. Each decision costs a prototype fee or a pilot fee, and each one comes before an institution has integrated anything, trained anyone or committed to an operating model.
Note
Why the counts are ten and five
They are one example of a shape, chosen because they show the ratio. A portfolio of six candidate workflows runs the same structure with smaller numbers. Selection still happens twice, both times on test results, and the cheaper stage still does the filtering.
The calendar
Eighteen weeks of delivery to a production decision
Modelled at five workflows running side by side. These are delivery weeks, which White Glove Labs contracts and is held to, measured from the opening workshop. Time to reach a contract sits outside them and belongs to you.
Show the detail
Delivery weeks measured from the opening workshop. The AI roadmap runs weeks zero to four. Prototypes one to five run weeks four to eight. Prototypes six to ten run weeks eight to twelve. Pilots one to five run weeks twelve to eighteen. A production decision falls at week eighteen. The production release and Managed AgentOps follow after that decision, are priced separately, and are drawn with dashed outlines.
Four weeks of roadmap. Eight weeks of prototypes in two cohorts of five. Six weeks of pilots. The production decision falls at week eighteen. Prototypes one to five finish at week eight and their results wait until week twelve, because we select across all ten together.
When does a program have to start to finish inside a fiscal year?
Eighteen weeks of delivery before a 31 March year end means the opening workshop falls in the last week of November. Reaching a contract is in addition to that and is yours to estimate, because you own the route and the review load. The reviewer questions page describes the routes available to a federal buyer.
Note
What sits after week eighteen
We scope the production release separately and publish no figure for it. Hardening, integration, the authority to operate, cutover and training are real work with a real cost, and none of it is inside the eighteen weeks. Managed AgentOps begins after a production decision and runs monthly on a six-month minimum, with no exit or handover fee after the minimum term.
The commercial shape
The first commitment does not grow with the program
A buyer running one workflow and a buyer running twenty write the same first cheque. That is what lets an institution start a program it has not yet decided to finish.
Show the detail
This page shows no program fee. No program price is true before we know the portfolio. We price larger work as fixed-price builds, scoped after the Diagnostic. The services page lists every engagement with its duration and its closing decision. We set a program value against a real list of candidate workflows and confirm it in writing.
What am I committing to when a program starts?
One engagement. The roadmap ends with a ranked sequence and a costed plan, and an institution that reads it and stops owes nothing further. The same is true at every later seam: each prototype and each pilot closes with a written decision, and you continue only by deciding to.
| Exit point | What exists by then | Owed after stopping |
|---|---|---|
| After the roadmap | A ranked, costed sequence of candidate workflows with control requirements against each | Nothing further |
| After the prototype cohorts | Ten working builds on your material, each with a written boundary and a recommendation | Nothing further |
| After the pilots | Five acceptance reports tested against cases with known answers, and one production decision | Nothing further |
| After a production decision | An approved system and the evidence behind the approval | Nothing, unless the release and operations are contracted |
How this is contracted
One award, delivered with a partner bench
A program bought as a sequence of separate awards spends more time in procurement than in delivery. The instrument matters as much as the method.
Show the detail
- One instrument with optional phases
- Where a buyer expects to continue past the first engagement, the requirement states the anticipated program value and the buyer contracts the whole program as one instrument with optional phases. A written decision exercises each phase. One award then covers the whole program.
- Contracted through a partner prime for federal work
- Standing partnerships exist with Canadian primes holding federal supply arrangements. We confirm the prime and the division of work in writing against the requirement before the buyer signs the roadmap. We contract provincial work and work for regulated firms directly.
- Delivered from a partner bench, by design
- At two people per workflow, five workflows at a time is about ten people at peak against three founders. Roughly seven of the ten come from partner benches, named in the contract with their role and their reporting line. A program of this size is a partnered delivery and is described that way in the statement of work.
- Not-to-exceed figures, set per engagement
- A not-to-exceed amount is set for each engagement in its statement of work before it begins, so the ceiling binds as a contract term.
What will my security, privacy and procurement colleagues ask?
The same questions they ask of a single engagement, because a program is a sequence of them. Where does the system run? Who can reach the data? What happens when an automated decision affects a person? Does the interface meet accessibility requirements? What happens at the end of the contract? The reviewer questions page writes out those answers, and they do not change with program size.
Before you plan around this
What this page does not establish
Show the detail
- We have modelled this shape and not delivered it. We built the eighteen-week schedule from durations White Glove Labs contracts for single engagements. No program of this size has run from roadmap through to a production decision at this firm, which incorporated on 27 September 2025.
- The enterprise scale depends on partners: roughly seven of ten people at peak come from partner benches. The contracting section above states this as a delivery fact a buyer has to plan around.
- We hold Designated Organization Screening at Reliability Status, so screened people can work on Protected material in your environment. We hold no Document Safeguarding Capability, so Protected material stays inside the client environment and a cleared facility is a partner prime’s to provide. On a program of this size that shapes where people sit more than anything else here.
- The work between a production decision and a live system is not on this page. We scope hardening, integration, the authority to operate, cutover and training separately and publish no figure for them.
- Ten prototypes and five pilots are an example of a shape. The counts show a ratio. The roadmap produces the numbers for a real portfolio.
Next step
Bring the list of candidate workflows
The call runs 30 minutes and costs nothing. Bring the workflows you are considering, and we will say which ones a roadmap would rank and which ones are not worth ranking.
If there are no candidate workflows yet, the AI roadmap is the engagement that assembles and ranks them. If a few candidates exist, the governance sprint answers which one to build, or whether to build a chosen one. It stands on its own. To see something working before anyone funds a pilot, the governed prototype runs on your own material.