Methods
Published evidence helps identify suitable approaches and what has already been established. Your business case still needs evidence of fit, quality, effort, benefits and full costs in your own work. Two connected methods help decide what to test, extend, improve or stop: an innovation process and a way of working with AI agents when the work involves them.
Part 1 · The innovation process
Each stage asks for only the evidence needed for the next commitment, and ends in a decision to go on, change or stop.
Challenge
Senior leaders set the important problems, opportunities and boundaries.
Spark
Turn the idea into something you can test: what should improve, for whom, against which baseline, and how a small test could show it. Consider simplifying or removing work before adding AI.
Test
Agree the measure and meaningful result before the test. Compare accepted output, time, cost and checking effort with the current process and, where practical, the best feasible alternative. Include ordinary and difficult inputs.
Scale
Decide whether the test justifies wider use. Record the eligible sites, assets or teams and how many are actively using the system. Track accepted work, checking, support and whether benefits exceed the full costs of delivery.
Operate
Accept the change into normal operation with a named owner and support route. Record when active use was last confirmed. Review quality, exceptions, costs and benefits at agreed intervals to check whether the expected value and controls persist.
Agree the measure and what counts as a good result before each test. Where a direct comparison is weak, use a parallel run or comparable work, and say what it cannot prove. Stopping is not failure: a good process stops weak or mistimed ideas before they absorb more time and money.
Who makes it work
People across the business
The people closest to clients and projects spot the waste and delay a central team misses. The process gives them a route to act.
Innovation Committee
A few senior people agree the challenges and make the back, stop, hold and scale decisions.
Innovation Management Office
A small team helps shape ideas, organises the evidence and keeps actions moving. It runs the process; it does not own every idea.
Each challenge needs a sponsor who owns the intended benefit. This can be small: a few clear challenges, a visible route for ideas and a regular decision meeting.
Part 2 · Working with agents
AI agents can carry out connected tasks using tools and information. 38thMove expects them to change how professional work is delivered and how firms organise their expertise. This method sits inside the innovation process: when a Spark or Test involves agents, it shows how to redesign the work, build and prove it, and decide who gains. Build one workflow at a time, and let each one inform the next.
1 · Choose the workflow
Pick a recurring task with a named owner, a checkable output and a clear standard for acceptance. For professional-services firms, bids, reports, planning and environmental evidence, project controls and knowledge search offer possible starting points. Each needs a defensible baseline and acceptance test. Engineering calculations and model changes need explicit validation first.
2 · Where to look first
For professional services, a useful place to start is knowledge work that you already buy in or budget for, that repeats and that is easy to check. Judgement-heavy work, such as trusted advice or sign-off, is last to hand over: use AI there for preparation and comparison. Test three areas separately: back office, existing client work made more efficient, and new AI-enabled services.
3 · Redesign first
Map what really happens, remove low-value work, simplify the handoffs, then choose the simplest response that works: a simpler process, ordinary automation, an assistant, an agent or a bought product. If an agent would only make waste run faster, stop.
4 · Build small and fast
A workflow owner, a technical builder and a working user shadow the real work and build alongside it. They start on dummy or masked data with limited access, then move to a controlled live test, then supported operation. Each step up is decided on the evidence, not by simply copying the code across. People stay accountable for consequential decisions and professional sign-off.
5 · Capture the expertise
Experts define the method, the checks and when to ask for help. That becomes reusable skills that survive a change of model or tool. Experts review corrections and rejected outputs, then test proposed changes before adding them to the reusable method. A shared set of approved components, such as access, testing and logging, lets later workflows start faster.
6 · Decide who gains
Released time is not a benefit until it is used. Decide early whether it becomes capacity, quality, margin, a lower fee or a new service, and who funds the shared capability. Changes in fees determine how the benefit is shared between the firm and its clients. New services need evidence that clients will pay.
The firm this points towards. Nearly all effort stays on today’s work. A very small group builds the shared platform, reusable skills and support for tomorrow’s. An even smaller group watches for changes that could reshape the business. This is a design principle, not a staffing target. Controls scale with the consequence of the work: light for a safe prototype, strong for live client work where people rely on the output.
Both methods are 38thMove's current working approaches, adapted to the size, risk and consequence of the work. They are not validated standards or a guarantee of productivity gains. The example workflows are illustrations, not measured results.