Fund the work
Choose which project, team or individual allowance pays. Keep the decision visible in the cost record.
Flowstate Helm
Know what AI costs. See what you get back.
One place to understand your AI investment. Fund your people and agents, connect spending to the work and find where it can do more for your business.
People using AI18,400
Connected tools6
AI spend$842,000
Recorded sessions by tool · selected groups
Your AI operating system Costs · Budgets · Adoption · Value
Your tools. Your team.
Keep the AI tools your team relies on.
Keep using ChatGPT and Claude in their own apps, with the features available on your plans. Your people can use the tools that help them do their best work.
Flowstate Helm brings connected usage, spending and budgets into one view. Help people get more from AI while keeping finance in control. The reporting and controls available depend on each connection.
See how your AI tools connect ↗01 / Spending
Catch rising costs before you overspend.
Bring subscription and usage costs from connected providers together. Follow them by person, team, project, model or provider.
Review changes over time, compare spending with the budget and see the forecast. Keep the basis for each figure clear, including costs that still need allocating.
Understand AI spending and value ↗ Customer support / September Forecast
£84,000 over budget at the current rate
240,000 replies this month
× £1.60 per reply
= £384,000 forecast spend
The cost per reply is driving the overspend.
Illustrative full-month comparison at the same reply volume. Quality review pending.
02 / A shared AI wallet
One budget. Across your AI tools.
Give a person or team a shared allowance across connected services. Fund project work from its own budget first, with rules for what happens when that funding runs out.
Allocate costs to departments, projects, funds or clients. Handle requests for more budget, alert the right owner and apply limits through supported connections.
Share a wallet across your tools ↗ Shared AI wallet This month
Budget £800,000
Used £624,000
Available £176,000
Spending by provider
Claude £360,000
OpenAI £180,000 Vertex £84,000
Next charge · Claude £4,800
Reporting pays £4,800
Project: £32,000 → £27,200 remaining. Team wallet: £176,000.
Approve more budget
Give requests an owner and a decision, with the reason for the increase alongside them.
Respond before overspending
Set alerts and supported spending limits. Decide which work should continue and which needs approval.
Across your business
A shared view for a better decision.
Finance needs to explain the bill. People need access to the right tools, whether they work in sales, finance, marketing or engineering. Team leaders need to know whether the work is improving.
Flowstate Helm brings those questions together, so the next investment starts with evidence.
03 / Adoption and waste
Keep useful access. Stop paying for the rest.
See who uses AI, where usage has lapsed and which tools teams rely on. Compare paid seats with observed use and find overlapping subscriptions.
Auto-suspend unused access under the rules you agree. Keep former employees, inactivity, renewals and plan changes in the same review.
Review tools, seats and renewals ↗Help people get more from AI.
Usage tells you who’s active. F.AI helps assess capability, task performance and frustration so you can see where support or a different model could help.
Explore F.AI ↗ Unused access Ready for review
1,800 seats Former employees and people no longer using their access
Claude 1200 unused seats · £20 per seat/month
£288,000 potential saving / year ChatGPT 600 unused seats · £24 per seat/month
£172,800 potential saving / year Potential annual saving £460,800
04 / Value
Compare the cost with the result.
Choose a task and agree the baseline. Measure preparation, review, corrections, output quality and running cost before and after AI support.
See whether the change creates capacity, reduces operating cost or improves delivery. Keep measured results separate from estimates and potential savings.
Create more capacity for your team ↗ Reporting across the business Before / With AI
Preparation
800h → 200h
Human review
100h → 100h
Total per cycle
900h → 300h
600h
Released for analysis and decisions each month.
£1,800 running cost per month 7,200 hours a year at 12 cycles
05 / Business activity
Follow AI use into the work it supports.
A provider bill doesn’t tell you which project benefited. Connect usage with project records, workflows and delivery evidence to understand the activity behind the cost.
Attribute AI use to business activity ↗The tool, model and recorded use
Start with the usage and cost records available from each connection.
The person, task and project
Use the records in your work systems to allocate cost and establish context.
What reached the business
Connect recorded AI output to delivered work. For code, compare it with GitHub changes and keep edits and evidence gaps visible.
Trace AI contribution to code ↗Engineering
See what AI contributes to the code you ship.
Follow AI assistance into pull requests. See which generated lines survive review, compare cycle time and inspect work that needed another pass.
Use the same approach across the company: connect AI activity to the work, then check the result.
Explore AI code contribution ↗ From prompt to pull request Illustrative data · Flowstate Helm
60%
of merged code came from AI
See the code attribution
1
export async function approveBudget(request) { Human 2
const budget = await getBudget(request.teamId) AI 3
if (budget.remaining > request.amount) { Removed 4
if (!canApprove(request.user, budget)) return Human 5
await reserveFunds(budget.id, request.amount) AI 6
return { approved: true } Removed 7
return recordApproval(request, budget) Human 8
} Human AI kept
Human written
AI removed
240 AI lines generated 96 kept
144 removed or replaced + 64 human lines
2.5:1 AI lines generated to kept
6 Prompts across 3 attempts
160 Lines in the merged change
PRs with AI assistance64 / 100
Merged PRs with linked AI evidence
Generated lines retained40%
96 of 240 generated lines made it into merged code
Cycle time comparison18h / 26h
Median open-to-merge: AI-assisted / no linked AI
Reopened PRs4 / 64
AI-assisted PRs reopened before merge
Repositories and quality checks
Repositories by AI usage
Linked coding sessions, split by tool
06 / Controls and evidence
Set the rules. Keep the evidence.
Choose models for the task
Compare cost and quality for coding, analysis or content work. Use optional model routing for supported tools and applications. Your team can keep using native apps alongside it.
Give the right people control
Set access around roles. Make budget ownership, approvals and exceptions clear.
Know where the figures came from
Trace costs to their sources, reconcile totals and distinguish measured spend from estimates.
Work with your privacy requirements
Agree what information the job needs, who can see it and how long it should be kept.
Workloads and agents
Follow the AI running inside your business.
Workloads show AI deployed inside applications. Agents show AI carrying out jobs. Give each an owner, a budget and measures that fit the work.
Understand workloads and agents ↗See agents next to the people they work with.
Flowstate Workforce helps you understand and plan agents alongside your people. See current commitments, match skills and availability to demand, then decide what AI can take on.
Explore Flowstate Workforce ↗Measure the useful result
For an application, compare cost per successful result. For an agent, follow completed jobs, exceptions and human review. Keep both alongside your company’s wider AI investment.
Your connected providers
Start with the tools your business uses.
We agree the reporting, attribution and controls available for your accounts. You can use Flowstate Helm on its own, or with Flowstate Workforce to understand and plan people and AI together.
Explore all integrations ↗
Make your next AI decision with a clearer view.
Request a demo ↗Practical questions
What to know before you start.
Can our team keep using ChatGPT and Claude?
Yes. Your team can keep using ChatGPT and Claude in their own apps, with the features available on your provider plans. Flowstate Helm works with your connected tools to understand usage, spending and budgets. You don’t need to move everyone into a new chat interface. Reporting and controls depend on each connection.
Do we have to use an AI gateway?
No. Flowstate Router is an optional gateway for supported tools and applications. It records usage and can route requests between models when you turn that option on. Native ChatGPT and Claude apps can remain part of your team’s work alongside it. We agree the right connections and controls for the tools you use.
Is Flowstate Helm only for engineering teams?
No. Flowstate Helm brings AI spending, budgets, adoption and activity together across the company. Finance, sales, legal, marketing, operations and engineering can each see the tools and work relevant to them.
Can Flowstate Helm stop every provider from overspending?
Controls depend on the connection. Some support spending limits or routing, while others provide delayed usage reports. We agree which measures and controls are available before setup.
How do you distinguish savings from unused budget?
Unused budget is headroom. Avoiding another purchase is cost avoidance. A reduced bill is a realised saving. We keep those measures separate so an estimate is not presented as cash saved.