Where to find it | Sidebar → Innovation → AI Solutions |
Who can see it | Leadership |
Can you change data here? | Yes — register solutions and take snapshots |
How fresh is it? | Client health refreshes weekly; business metrics recompute live |
What it's for
Our AI/ML team ships tools into specific parts of the business, and leadership needs evidence rather than anecdotes that they worked. Without measurement discipline, a tool launches, everyone moves on, and months later nobody can say whether revenue per employee, retention or client health actually changed for the teams that got it.
This page closes that gap in three steps:
Register the deployment on the day it goes live — what it is, and who got it: a division, a department, one person, or a set of people.
Freeze a baseline — business metrics and client-health scores exactly as they stood at deployment.
Watch the deltas — current values recompute continuously, and client health is snapshotted every week, so improvement or degradation against the baseline is always visible, with the deployment date marked on the trend chart.
It's a longitudinal ledger for AI investment, feeding the decision about which solutions to scale, pause or kill.
What you'll see
The solutions list — every registered deployment with its type, status, and a health column showing current average churn risk with the change since baseline.
Open any solution for:
Baseline versus current business metrics — for an org target: total revenue, average RPE, active clients, seats, retention and employees. For a person target: attributed revenue, MRR and RPE.
Client health, baseline versus current:
Metric | What it means |
Avg Churn Risk | Average risk score across the tracked clients |
% At Risk | Share above the risk threshold |
Sentiment Risk | From calls and check-ins |
Avg SE Readiness | Expansion readiness |
Avg Retention Score | Model retention score |
Clients Tracked | How many clients are in scope |
A weekly health trend chart with churn risk and expansion readiness lines, and the deployment date drawn on it — which is the whole point. It defaults to eight weeks before deployment through today.
A revenue trend chart across snapshots, a per-person breakdown for multi-person targets, and a snapshot history recording what was captured, from where, and by whom.
How to use it
Register on launch day, not later. Create the solution, pick the target, and take the baseline. This is the step that can't be recovered afterwards — a baseline captured three months in isn't a baseline. When you pick a target, the form previews roughly how many clients will be health-tracked.
Review monthly. Open the solution and read baseline against current. The business metrics answer "did output change?"; the client-health metrics answer "did the clients feel it?"
Read the trend chart, not just the deltas. The deployment marker is there so you can see whether the line bent at deployment or was already moving. A number that was improving before you shipped isn't your win.
Decide. Scale what moved, pause what didn't, and use the snapshot history as the evidence trail.
Good to know
The baseline is the whole value of this page. Register the deployment when it happens. A baseline can be approximated from historical model scores if it's missing, and the page tells you when it did that — but an approximation is weaker evidence.
Snapshot provenance is shown, and worth reading. Each figure says whether it was frozen at baseline capture, reconstructed from a specific week, or approximated from history near deployment. Weight your conclusions accordingly.
Business metrics are monthly-grained, and captured independently of the app's month selector. If data isn't available at the exact scope requested, Signal falls back to a broader scope and records that it did — visible as a badge.
"Employees" on an org target is derived, not counted. It's inferred from revenue and RPE, and flagged as derived in the capture record.
The per-person breakdown uses current team assignments, not the assignments as they were at deployment. So a person's book may have changed underneath the comparison.
Sentiment only counts clients with a real sentiment source. Model defaults are excluded rather than averaged in, which is why the client count for sentiment can be lower than clients tracked.
Health tracking excludes churned and contract-ended clients, so the tracked set shrinks over time as clients leave.
Client health snapshots run weekly. New points appear once a week, not continuously.
The revenue trend chart only gains a point when someone takes a snapshot or a newer data month lands. Weekly health snapshots don't add points to it.
Common questions
Who can see this page? Leadership. See Who can see what.
Can I add a baseline after the fact? Signal can approximate one from historical model scores near the deployment date, and it labels it as approximated. It's better than nothing and worse than a real baseline. Register on launch day.
Why is the tracked client count falling? Churned and contract-ended clients drop out of health tracking.
Why is the sentiment client count lower than clients tracked? Only clients with a genuine sentiment source are counted. Model defaults are excluded deliberately.
Why don't the revenue figures match the finance data exactly? They're captured for a specific month and scope, possibly with a documented fallback to a broader scope, and allocated across team assignments. The capture record tells you exactly what was measured.
The employee count looks wrong. On org targets it's derived from revenue and RPE rather than counted, and flagged as such.
Why did the trend chart barely change after deployment? That may be the finding. Check whether the line was already moving before the deployment marker — the marker exists precisely to prevent crediting a pre-existing trend.
Related
Company Capacity — where to deploy AI next, and its own local deployment log
Churn Risk — the source of the health scores here
Service Expansion — the source of the readiness scores
Employees — the per-person revenue behind employee targets
Glossary — RPE, churn score, expansion score