Appearance
How Flowstate fits together
Every part of Flowstate reads from the same records: your people, contractors and AI agents, the teams and projects they work on, and the tools that record the work. Read this before you plan your set-up: what you set up once feeds every screen, and a gap in one place shows up in several.
The records everything depends on
| Record | What it holds | Where you set it up |
|---|---|---|
| Organisation settings | Reporting currency and exchange rates, fiscal year, locations, resource types (overhead), job roles, and cost centres with their category (CapEx, OpEx, R&D) | Organisation basics |
| People | Employees with their salary history and bonuses, contractors with their rates, planned hires (vacancies) and AI agents | Employees, Contractors, Vacancies, Agents |
| Teams | Your team hierarchy, and each team's manager, who needs a Flowstate account | Teams |
| Allocations | Who is planned on which team and project, from when, and for what share of their time | Allocations |
| Projects and initiatives | Each project's owner, dates, cost centre and drivers. Initiatives group projects together. | Roadmap |
People and AI on one plan
Flowstate treats AI as part of your workforce, beside employees and contractors, and places its cost against the same teams and projects.
- Copilots are AI tools people use, such as Claude Code or ChatGPT. Their cost belongs to the person using them.
- Agents are AI workers you plan in Resourcing → Agents & copilots, with an owner, on projects, with their own cost.
- Hybrid workforce puts employees, contractors, and agents and copilots on one bill.
See The AI cost model.
From AI adoption to what it returns
Each question builds on the one before, and each has its own screen.
| Question | Where you answer it | Built from |
|---|---|---|
| Are people using AI? | Adoption | AI sessions, the seats your providers bill for, and who is on which team |
| Where does AI sit in the work? | Agent sessions and AI impact | AI sessions tied to projects and pull requests, and the agents you run |
| What does the work return? | Business metrics | What a team produced, beside what its people and AI cost, with and without a person |
| Did the project deliver? | Project value | A project's drivers read before and after the work, beside what it cost |
A better result frees up capacity. It only becomes a saving when headcount comes down or the team takes on more work.
Two ways Flowstate works out cost
Both start from the same day rate for each person. The formulas are in How costs are calculated.
| Forecast: what people are planned to cost | Actual cost: what the work actually cost | |
|---|---|---|
| Comes from | Allocations: people, contractors, vacancies and AI agents on teams and projects | Effort: the share of each day people spent on each project, estimated from your tools and submitted by team leads |
| Looks | Forward, across the fiscal year | Back, week by week |
| Feeds | Forecast, budgets, scenarios, headcount analytics | Capitalisation, R&D tax claims, the Spend timeline, initiative actuals |
A budget saves a copy of the forecast as the year's budget. An accepted budget run also gives each team its own budget, but a snapshot doesn't. Capitalisation and R&D claims never use the forecast. They only use effort that has actually been worked.
Engineering insights prices work a third way. The cost of a pull request comes from who committed to it each day and what they cost per day. See Pull requests and tickets.
Where effort comes from
Once a day, Flowstate estimates effort from ticket activity in your project tool, sharpened by GitHub if it's connected. Team leads then check and submit each week, and finance uses the submitted weeks. See How the effort estimate is built.
What each part needs first
| Part | Reads | Gives you | Needs first |
|---|---|---|---|
| Forecast | Allocations, salaries, rates, vacancies, overhead | Planned cost by project, initiative, team or cost centre | People, teams, allocations, fiscal year, exchange rates |
| Scenarios | A private copy of your live records | Proposed changes, reviewed and merged into live data | Live records |
| Budgets | The forecast and team managers | Budget versions for the year, and a budget for each team when a run is accepted | Forecast, a manager on every budget-holding team, the budget workflow |
| Effort reporting | Project tool, GitHub, team managers | A submitted week for each team | Project tool connected, people matched, projects linked, weekly review turned on |
| Capitalisation | Submitted effort and day rates, each project's cost centre, your CapEx framework | The CapEx/OpEx split, a justification for each project, completed capitalisations and exports | Effort reporting, salaries and rates, categorised cost centres, a default framework, review stages |
| R&D tax claims | Effort cost, with tickets and pull requests as evidence, and your R&D framework | Suggested claims, claim documents, claim packs | Effort reporting, salaries and rates, a default R&D framework, the eligibility scan |
| Engineering insights | GitHub pull requests and commits, tickets that tie them to projects, AI sessions | Cost per pull request, flow times, AI impact | GitHub connected and contributors matched, and a project tool for project and initiative links |
| AI spend | Your AI providers' billing | Spend by person, team and tool | Provider connections, and people with the email they use at each provider |
| Adoption | AI sessions, provider seats, team allocations | Who uses AI, how deeply, which teams are behind, and seats nobody opens | People on teams, the Cloud Proxy, provider connections |
| Agent sessions | AI sessions from the Cloud Proxy | Sessions tied to people, projects and pull requests, and spend by project and kind of work | The Cloud Proxy installed on people's Macs, and projects linked to your project tool |
| Business metrics | Readings of what a team produced, and what its people and AI cost | Cost per unit, and how it differs when AI does the work instead of a person | Salaries and rates, AI providers connected, a metric with readings |
| Project value | Drivers on a project, their readings, the project's cost, and the business metric its team produces | Whether a project moved what it set out to, beside what it cost to build | Driver types, drivers on projects, readings before and after the work |
| Hybrid workforce | Salaries, rates and AI spend | Employees, contractors, and agents and copilots on one bill | Salaries and rates, AI providers connected |
| Functional groups | People | Your functional structure, positions and their cost | People |
What goes wrong when something is missing
| Missing | What you'll see |
|---|---|
| A salary or rate | That person's effort has no cost ("uncosted effort"), so CapEx and R&D figures come out too low |
| An exchange rate | Amounts in that currency aren't converted. They're counted as if they were already in your reporting currency |
| A team manager | The team gets no budget proposal, and nobody is asked to submit its effort |
| A project's link to your project tool | Its work lands in Unattributed, and the week can't be approved until that's sorted |
| A CapEx cost centre on a project | Its cost isn't capitalised |
| A weekly review cadence | It's off until you turn it on, so team leads aren't asked to submit their weeks. Views that count only submitted effort stay empty |
| A default CapEx framework | No justification documents are created |
| A default R&D framework | The eligibility scan doesn't run |
| GitHub contributors matched to people | Their code doesn't shape effort |
| An employee's team | Their AI use isn't counted in any team on Adoption |
| The Cloud Proxy on people's Macs | Adoption shows nobody as active, sessions and spend by project stay empty, and AI impact shows only agent-written pull requests |
| Whose cost this is on a business metric | The metric is counted but not priced |
| A driver's reading before or after the work | The driver shows Not read |