Make one workflow work properly, on your own information
Typically four to six weeks to build one working workflow on your organisation's own information, to the standard production use demands: sources people can check, the right permissions, expert review, evaluation and before-and-after measurement.
Your people use AI every day. Has the work changed?
Most organisations now have AI tools and a few pilots, and very little that has changed how the work gets done. A demonstration on sample documents proves little. The hard parts are your own information, who may see what, how an expert checks the output and how anyone knows the work is better.
The sprint takes one workflow, usually chosen in an AI Review, and builds it properly. We baseline how it works today, build it on your own information with citations, permissions and expert review designed in, benchmark it against agreed tests and measure it against the baseline.
It ends with the workflow in use or ready to go live, the integration requirements written down and a costed pathway into production, so the next decision rests on evidence from your own work.
How a sprint runs
One workflow on your own information, built and measured in five phases
Typically four to six weeks from the baseline to the evidence.
Work through one workflow first with the free AI Brief TemplateBaseline
Map how the workflow runs today and record the time, quality, rework and cost before anything changes.
Build
Build the workflow on your own information, with citations people can check and permissions that follow who may see what.
Benchmark
Test it against an agreed evaluation set, so quality is measured rather than assumed, and fix what the tests find.
Expert review
The people who own the work review the output, agree where their judgement stays in the loop and approve it for real use.
Measure
Measure the workflow against the baseline, separate what has been shown from what still needs evidence and set out the production pathway.
Adoption
Capability built through the work
The team learns to delegate to AI, review what it produces, recognise when it should not decide, work with agents and keep a person accountable for the result, all inside its own workflows.
Use
Get reliable results from the tools the team already has.
Judge
Check AI output against its sources and know when it should not decide.
Design
Redesign a workflow around what people and AI each do well.
Delegate
Hand work to AI and agents with clear instructions and limits, and a named person accountable.
Who this is built for
- Research, insight and advisory firms whose experts spend hours finding, checking and re-creating knowledge
- Leaders whose people use AI tools but whose results have not moved
- Organisations that have finished an AI Review or Discovery and want to act on the first opportunity
- Teams that need AI output they can trust with client or regulated work
What you receive
- A baseline of the workflow today: time, quality, rework and cost
- One working workflow on your own information, with sources people can check
- Permissions that respect who may see what, and an expert review step wherever judgement matters
- An evaluation set and its results, before and after each change
- Before-and-after measurement against the baseline
- The integration requirements and a costed pathway into production
Typical investment
Fixed scope, fixed price
Know what the answer costs before we start. All figures ex VAT; the final fixed fee is confirmed after scoping.
- One workflow on your own information
- from £25,000
- Larger collections, complex permissions or deeper integration
- scoped separately
What we need from you
- The experts who own the workflow, with time set aside to review and test it
- Access to the information, systems and permissions the workflow depends on
- Someone who can approve the workflow going live
Scope and boundaries
- The volume and state of the information, the permission model and the depth of integration set the price. Not every sprint costs £25,000.
- It works on your own information and your own workflow. It is not a generic AI training course or a demonstration on sample data.
- Software licences and cloud consumption are separate, and are only recommended where the evidence supports them.
- Taking the workflow across a function is an AI Programme. Ongoing evaluation and improvement is Managed Improvement.
Sample extract
What you get
An illustrative extract, not client data. The real document carries your evidence and your numbers.
TechGuidr
Sample extractAn extract from a sprint results note
- Workflow
- Preparing a client briefing from previous research and current data
- Baseline
- Analyst time per briefing and corrections after expert review, recorded before any change
- Build
- Drafts from the firm's own publications and datasets, each point citing its source, with analysts seeing only what they are permitted to see
- Stays with people
- The judgement on what the client needs, every figure and the decision to send
- Result
- Measured against the baseline, with anything not yet shown marked as such
Our own product
Relevant work
Radicand runs a production workflow across a research business’s proprietary knowledge and structured data: more than 78,500 publications, 25 years of research and its forecasts, with sources shown so each claim can be checked. A Sprint brings the same discipline to one workflow of yours, whether or not Radicand is part of it.
See RadicandThe AI workflow ladder
Where this sits
Start with a review, prove one workflow on your own information, then scale what works and keep it improving.
AI Review
from £7,50010 working days
Ten working days on how the work really happens: where expert time and value leak, the top three opportunities, how ready the knowledge and data are, the controls needed, the economics, an architecture direction and a prioritised roadmap.
AI Sprint
from £25,000Typically 4 to 6 weeks
One working workflow, built on your own information: citations, permissions, expert review, evaluation, before-and-after measurement, the integration it needs and a pathway into production.
AI Programme
from £60,000Scoped to the function
Scale a proven workflow across a function or the business: further workflows on shared foundations for knowledge, permissions and evaluation, with the adoption work that makes the change stick.
Managed Improvement
from £5,000Per month
Keep the workflows in use getting better: ongoing evaluation, new workflows, model changes, optimisation, governance and improvement on an agreed monthly remit.
Smaller business or founder? Start with an AI Brief session: 90 minutes with Dan and a written report, £1,250. Prices exclude VAT.
Before you commit
Questions before you commit
How is this different from the AI Review?
The review spends ten working days finding where expert time and value leak and which workflows to change first. The sprint takes one of them and makes it work properly on your own information, then measures the result.
How do you make the answers trustworthy?
Every answer is grounded in your own sources and cites them, so an expert can check it. Permissions decide what each person can see, an expert review step sits wherever judgement matters, and an evaluation set tests quality before and after each change.
Do we need to buy new software?
Not at the outset. We start by assessing the tools and platforms you already have. Where something new is worth buying, we say so, with the cost and the alternatives, and there is no resale margin behind the recommendation.
What happens after the sprint?
You keep the workflow, the evaluation set, the measurements and the production pathway. The next step might be taking it into production, an AI Programme across the function, Managed Improvement, or nothing more from us.
How do you measure AI adoption and ROI?
The workflow is baselined before anything changes: time taken, quality, rework and cost. The same measures are taken after the change, and anything not yet shown is marked as such. No return is promised in advance.
Does it work with Microsoft Copilot or ChatGPT?
Yes. We start with the tools the team already has, such as Microsoft 365 Copilot or ChatGPT, and build beyond them only where the workflow needs sources, permissions or evaluation they cannot provide.
The route
Other services
Every engagement sits on the route from an idea to a working system. Start wherever you are.
Talk through a sprint for your workflow
Two or three sentences is enough. We will tell you straight whether and how we can help.