Explore Flowstate

The Flowstate platform
Helm

Know what AI costs. See what you get back.

Understand spending, share budgets and connect AI use to the work it supports.

Recent insights All insights

Buy tokens, rent GPUs, or own the rack?

Open-weight models are now good enough to make self-hosting a financial question rather than a capability one. I modelled four options, an office rack, colocation, rented GPUs and per-token APIs, to find where each one actually wins. The answer turns on utilisation, and it isn't the one the internet keeps giving.

Turn a $3m AI bill into $1.9m

Most teams overpay for AI twice without noticing: they run everything through the default flagship model, and they buy the tokens on a metered plan they can't see into. An intelligent proxy fixes both. Here is the maths, a calculator you can drag, and where Flowstate fits.

Start with the work All use cases

Understand AI spending and value

See where the budget goes and what the business gets back.

Share one AI budget across your tools

Fund people, teams and projects across connected services.

Give your team more capacity

Reduce the preparation, checking and follow-up around recurring work.

Connect AI use to delivered work

Follow AI activity into projects, code and business results.
Quick links
Flowstate / People + AI Talk through your priorities
Use cases

See which AI-generated code reaches your codebase.

Connect AI generation records with GitHub to trace output from tools such as Codex, Gemini and Claude into committed code. Review what was kept, changed or removed across your connected tools.

Talk through your process

Connect output to the change it produced.

Use generation records and repository history together. Follow recorded output from the AI tool to a proposed change and its final merged version.

Explore AI impact on engineering ↗
Generation to delivery
ToolChangeResult
Codex Change 142 Merged with edits
Claude Change 143 Merged
Gemini Change 144 In review
Traceability needs both generation records and repository access

Separate generated code from the final result.

Compare recorded output with the version people reviewed and merged. Account for edits and removed code rather than counting everything the model produced.

Explore agents, data stores and views ↗
Change 142 · Matched lines
Recorded outputLines
Kept unchanged 128
Edited before merging 42
Removed before merging 31
201 recorded lines reviewed against the merged change

Make gaps visible.

When a tool doesn’t provide the required generation history, leave the origin unconfirmed. Use the evidence you have to assess contributions without labelling every commit as AI work.

Connect AI use to work and projects ↗
Attribution coverage
EvidenceAssessment
Generation and merge records Traceable contribution
Generation record only Delivery unconfirmed
Repository change only Origin unconfirmed
Code volume alone doesn’t establish value or quality
Your systems, connected

Use the tools behind this work.

We agree which records, access and controls your workflow needs, then connect the relevant systems.

Explore all integrations ↗
Claude Code Discuss setup OpenAI Platform Available Google Vertex AI Discuss setup GitHub Available

Measure what changes.

Track the share of changes with traceable generation records, matched code kept in merged changes and human edits. Keep unmatched code separate instead of guessing its origin.

We agree the baseline with you, build around your systems and keep improving the application as your work changes.

More work we can help with.

Finance reporting

More accurate reports. More time for your finance team.

 Deal approvals

Get complete deals to the right approver.

 Workforce planning

Turn the company budget into a workforce plan.

Start with the work you want to improve.

Show us the process, the systems and where your team loses time.

Talk to Flowstate
Practical questions

What to know before you start.

What is the difference between generated and retained lines?

Generated lines are what an AI tool wrote. Retained lines are the generated lines that survived into merged code. A count of additions alone cannot establish retention: it needs a link between the generated output and the final change.

Does a reopened PR prove that AI caused a problem?

No. A reopen flags work to investigate. Compare the original task, review and follow-up changes before drawing a conclusion. Cycle-time differences also need comparable groups rather than an automatic causal claim.