FlowCraft Labs runs founder-led working sessions with Business, Finance, Risk and Data teams, helping them build more consistent and reliable AI workflows across analysis, reporting and recurring work.
Better task selection. Better context. Stronger verification. Smarter use of skills, connectors, MCPs and agent orchestration. Clear lines on what stays human.
We're selecting 10 organizations for the founding FlowCraft AI Lab cohort.
AI adoption has moved faster than organizational practice.
People are experimenting individually. Useful approaches stay trapped with individuals. Output quality varies. Verification is often informal. And teams rarely have a shared way of deciding what should be delegated to AI, how the workflow should be designed, which capabilities should be connected, and where human judgment must remain central.
Everyone develops their own way of working with AI.
One analyst finds a better approach to a recurring task. Everyone else continues doing it differently. What works rarely compounds across the team.
Two analysts can give the same model the same task and produce very different results.
The difference is often context, instructions, process and verification — not simply the model itself.
Most teams stop at chat.
They rarely connect AI to the context, tools and repeatable processes that make it genuinely useful — through reusable skills, connectors, MCPs or coordinated agents.
AI accelerates weak analytical processes just as easily as strong ones.
Generating an answer faster is not the same as improving the quality of the decision behind it.
The more convincing AI output becomes, the easier weak reasoning is to miss.
For analytical teams, producing work faster is only half the problem. The harder question is knowing what to trust, what to check and what should never be delegated.
This isn't generic AI training.
The FlowCraft AI Lab applies practical principles directly to the analysis, reporting and recurring workflows your team already handles — including how to move beyond one-off prompting into more capable, repeatable AI workflows.
Build a practical mental model of what current AI systems do well, where they fail and why human judgment still matters.
Work through real analytical tasks using stronger approaches to task selection, context, instructions and verification.
Learn where skills, connectors and MCPs can give AI the right tools and context to operate inside a real workflow rather than an isolated chat.
Explore when a single AI interaction is enough, and when a workflow benefits from multiple steps, tools or agents working together.
Turn what your best people do individually into repeatable workflows the wider team can use.
Your team leaves with a shared method for deciding:
Before AI, producing the first draft of an analysis often consumed much of the work.
Increasingly, the bottleneck moves downstream:
As production gets cheaper, judgment and workflow design become more valuable.
Giving everyone access to AI is not the same as building AI capability.
Business analysts and operations teams producing requirements, process analysis, business cases, reporting and stakeholder recommendations.
FP&A, corporate finance and finance business partnering teams working across modeling, reconciliation, reporting, commentary and decision support.
Risk, compliance and internal audit functions where outputs need to be challenged, evidenced and understood — not simply generated.
Data scientists, analysts and analytics teams working across analysis, documentation, querying, quality assurance and technical decision-making.
Prompt quality matters, but it is only one layer of a useful AI workflow. FlowCraft teaches teams how to combine the right capabilities around the task.
Create reusable instructions, standards and expertise so good practice does not have to be rebuilt from scratch every time.
Bring relevant documents, systems and business context into the workflow so AI can work from the information the team actually relies on.
Connect AI to tools and structured capabilities that let it retrieve information or take controlled actions inside a wider workflow.
Design multi-step workflows where different AI actions, tools or agents handle distinct parts of a task — with human review at the points that matter.
Build explicit checks around AI output so speed does not come at the expense of analytical quality, accountability or decision-making.
The goal is not to use more technology. It is to design the simplest AI workflow that produces better work.
FlowCraft may teach persistent context, reusable skills, connectors, MCPs, agent orchestration and other practical capabilities.
But the method is not tied to whichever AI product or feature launched last month. The underlying questions are more durable:
That is the capability FlowCraft is interested in building.
FlowCraft Labs is being developed through direct work with analytical teams.
The first cohort is deliberately limited to 10 organizations so we can observe where AI is genuinely improving analytical work, where teams are struggling, which workflow patterns create the most value, and which problems keep appearing across Business, Finance, Risk and Data functions.
In return for participating in that research, selected organizations receive the founder-led Lab without charge.
The cohort closes once all ten suitable organizations have been selected.
Data scientist and published LLM researcher with experience across Google Research, Credit Karma, the UK's national statistics agency and the UK defense sector.
His work has included production machine-learning systems, ranking models, data quality, national economic statistics and research into large language model capabilities.
FlowCraft Labs brings that technical and analytical background into one practical question:
How should serious analytical teams design AI workflows when the quality of the output and the judgment behind it actually matter?
We're looking for Business, Finance, Risk and Data teams already experimenting with AI — or looking for a more systematic way to move from isolated prompting toward optimized AI workflows.
To apply, tell us:
Your team already has AI.
The next question is whether it has a good way of working with it.
Apply for a Lab Four short questions. We reply within two business days.10 organizations only · September–November 2026 · No fee for selected teams