Eight back-offices, one person switching between them.
10B+messages · founder's prior work
6warehouse domains · founder's prior work
239tech stacks audited · our tooling
⌐ 01
Five modules
All of them share the same machinery: a
verification queue, rules that accumulate, and a record of every step.
Document and spreadsheet agents
Files in, records out; a person confirms the uncertain.
What's included
Format recognition, column mapping, validation, write to the target system. Uncertain judgements queued for confirmation; confirmations become rules, so the next file from the same source is handled automatically.
Classification and extraction
Unstructured to structured: tickets, reviews, contracts, return reasons.
What's included
Taxonomies agreed with you; confidence thresholds; low-confidence items routed to a person. Output lands in the warehouse as structured fields.
Operations agent
One surface across back-offices; writes need confirmation.
What's included
Inventory, orders, finance, support in one view; every write action proposed by the agent, confirmed by a person, logged.
Ask your data
Plain-language questions, answered with the query shown.
What's included
A semantic layer over your definitions. Every answer shows the query it ran; no answer without a source. Runs on the Data Foundation.
Managed operations
Kept running as systems change. Monthly.
What's included
Interface-change tracking, queue and rule review, a monthly report of what ran, what was confirmed and what failed.
⌐ 02
How it works
Your dataFacts from systems of record
Agent proposesA judgement, with evidence
Person confirmsVerification queue
Execute and recordWritten to your systems, logged
Rules accumulateConfirmed judgements become rules
Step 05 feeds back into step 01
⌐ 03
What we connect
ERP
CRM
Support desks
Email
Spreadsheets & files
Shared drives
Your warehouse
The AI you already run in support and marketing stays; we make sure it sees the same data.
⌐ 04
Your data stays yours
Your environment
Read-only by default
Your model channel
In the contract
⌐ 05
Agents that ask when unsure. Rules that accumulate. Records of everything.
How this compares with RPA, AI copilots and building in-house
RPA
AI copilots in SaaS
Build in-house
Unvessel
Handles variation
Breaks on change
Within its product
If you build it
Judges; asks when unsure
When unsure
Fails
Guesses
Depends
Queue to a person; answer reused
Data in view
One screen at a time
Its own product
Depends
Every connected system
Verifiability
Logs
Black box
Depends
Every step recorded
Way in
Licence
Bundled
Hiring
A two-to-three-week diagnostic
⌐ 06
How to start
DiagnosticWhich routine work first; included in the Data Health Diagnostic
First workflowFixed price, accepted on its own
Managed operationsFrom go-live, monthly
We don't automate what the data can't support yet.
We don't let agents write without confirmation.
We don't replace your systems of record.
⌐ 07
Questions
Is this RPA?
No. RPA replays clicks and breaks when a layout changes. These agents read and judge, and ask a person when they can't.
Do we need the Data Foundation first?
For document agents on one system, no. For anything that spans systems or answers with numbers, yes.
What happens when the agent is wrong?
Uncertain judgements never execute; confirmed ones are on record and can be withdrawn. Effects are traceable.
How is it priced?
Diagnostic at a fixed fee; each workflow at a fixed price with acceptance criteria agreed first; managed operations monthly.
Tell us which routine work eats the most hours and which systems it touches. We'll reply with the scope and the fee — or tell you honestly if we're not the right firm.