Run the Retention Analyst

This demo isolates the reasoning layer, not the full system. Every analysis below runs live through the Claude API, no canned responses. In production, this step sits downstream of a data pipeline (usage metrics, CRM context) and upstream of delivery (Slack). What's being tested here is the part that matters most: can the AI actually look at the numbers and make a defensible call, not just flag a threshold and call it a day.

Pick a few accounts below, hit Run Analysis, and see for yourself. Every verdict cites the real numbers it's based on, so the reasoning can be checked against the source data rather than taken on faith.

The table below shows sample weekly API usage for a fictional SaaS company.

AccountWeek 1Week 2Week 3Week 4
Vela Studio900880910895
Solace Group1,0007002000
Northbridge Co1,1001,1501,3001,400
Ashgrove Ltd300280310290
Cobalt Ventures1,5001,6001,8502,000
Tern & Co150410
Driftwood Media750500220180
Quint Logistics400410395405
Brackish Media1,8001,7501,7201,680
Brightfield Partners2004508001,250
Petrel & Co38015000
Marlow & Co600580510470
Forsythe Partners5060451,400
Holloway Group600450700880
Halden Industries1,2001,1801,2201,300

Heads up: you can select up to 8 accounts and get 2 runs on this demo (it's calling a real AI model, not a canned response). Choose your accounts wisely.