Discovery first. Then a build that lasts.
Two phases and a standing commitment. We find the work worth doing, build it on a foundation that’s secure from day one, and then stay to run it — as your outsourced AI CTO.
Find the work worth doing first.
Before anything is built, we find where AI will actually pay back — by talking to the people who do the work, reading the systems it happens in, and testing what’s feasible before any heavy investment.
Where the hours go
Interviews with the people who do the work, and a close look at the reports, processes and questions that eat their week.
Systems and data review
What you run, where the data lives, who can see what — and which systems won’t let AI in directly.
Security and risk
How data will be protected, retained and audited, agreed with your IT and security leads up front.
Roadmap and business case
The first use cases ranked by hours saved and feasibility, with what they cost to build and to run.
- 1Monthly board packFinance & leadership180Data ready
- 2Supplier renewals reviewLegal & procurement95Needs staging area
- 3Weekly operations reportOperations60Data ready
- 4Support ticket triageSupport140Needs labelling
- 5Salary benchmarkingPeople20Need-to-know data
Ranked by hours saved and how ready the data is. The first two go into the build plan, with cost to build and run.
The output is a ranked, costed plan you can take to your board.
Build it on a foundation that’s secure from day one.
Then we build — in an order that doesn’t bend. Seats first, then the data, then the walls, and only then the workflows, so everything that ships inherits the controls instead of retrofitting them.
- 1 Claude Enterprise rollout
- 2 Data foundation
- 3 Guardrails, security & the harness
- 4 Workflows & agents
Security lands before any workflow goes live.
Claude Enterprise rollout
We stand up Claude Enterprise for the whole company: single sign-on, group provisioning, workspaces and projects shaped around how your teams actually work, connector approvals, a usage policy people read, and the first use cases — chosen with each team, not for it.
- SSO and group provisioning mapped to your org chart
- Workspaces and projects structured by team
- Approved connectors, everything else blocked
- Usage policy, training and first live use cases
- Single sign-onConnected
- Group provisioning6 groups synced
- Projects by team and project14 created
- Connectors5 live · 1 pending · 1 blocked
- Usage policy61 of 67 acknowledged
Seats live by team
- Leadership6 seats
- Sales18 seats
- Operations14 seats
- Legal5 seats
- People4 seats
- Support20 seats
Data foundation
Your data already exists — in documents, inboxes, the CRM, the ERP, the ticketing system. We connect it into a governed lake, add staging areas for systems that won’t allow AI in directly, and model it as an ontology: customers, contracts, products, suppliers, people and how they relate. AI works on your company’s real structure instead of a pile of files.
- Connectors into the systems you already run
- Staging areas for systems that won’t open up
- An ontology of your entities and their relationships
- Retrieval over documents and records, permissions intact
| Source | Carries | Status | Last sync |
|---|---|---|---|
| SharePoint documents | policies, contracts | Synced | |
| shared inboxes | Synced | ||
| CRM | customers, contacts | Synced | |
| ERP · staging area | orders, spend, suppliers | Synced | |
| Chat | team channels | Synced | |
| Ticketing | support tickets | Awaiting approval | |
| Personal file shares | unmanaged drives | Blocked by policy |
Every record keeps lineage to its source system.
Guardrails, security & the harness
Before a single workflow touches production, the controls are in: zero data retention on every model call, need-to-know walls around sensitive data, permissions mirrored from your systems, evals that catch wrong work before your team does, and a full audit trail. The harness everything runs inside is built here — so every later workflow inherits it.
- Zero data retention and need-to-know walls
- Permissions mirrored from every source system
- Eval suites run before each release
- Audit trail and observability across every call
- Need-to-know: People data stays in People workspacesEnforced
- Zero data retention on every model callEnforced
- Customer contracts: Legal and account owners onlyEnforced
- Answers must cite a sourceEnforced
- Irreversible actions need sign-offEnforced
- Evals pass before releasePassing
Caught in one morning · illustrative
- 08:14:02Redacted
Prompt pasted a customer’s personal details — redacted
- 08:31:47Blocked
Sales user asked for salary data held by People
- 09:02:15Held
Answer lacked a citation — returned for retrieval
- 09:40:26Held
Agent drafted a supplier cancellation — held for sign-off
Workflows & agents
Only now do we build the workflows: the board pack that takes a week, supplier renewals, support triage, contract review, the weekly report someone assembles from six tools. Each runs inside the harness, shows its sources, and hands a person the decisions that can’t be undone.
- Workflows chosen by the hours they cost, not the novelty
- Every output traced back to its source
- Sign-off wherever a step can’t be undone
- Measured against the time it used to take
- Read contracts expiring in 90 daysfrom Documents · 9 supplier contractsDone
- Pull spend by supplierfrom the ERP staging area, as of 06:00Done
- Check terms against policynotice periods, auto-renewal, citationsPassed
- Flag 2 for Legalnon-standard liability terms · Arden Supply +1Sent
- Draft renewal recommendationsone page per supplier, every figure citedWriting
- Operations lead sign-offthe decision stays with your teamWaiting
Then we stay, and run it.
Nothing gets handed over and left to rot. We operate what we build, keep it current as models improve, and extend it as more of the company comes on board. The data and the capability stay yours.
Your outsourced AI CTO
We operate what we build: monitoring, incident response, model upgrades tested against your evals before they ship, new connectors, capacity as more of the company logs in, and a quarterly roadmap reviewed with your leadership. The data and the capability stay yours throughout.
- Monitoring, on-call and incident response
- Model upgrades, tested against your evals first
- Capacity planning as usage grows
- A quarterly roadmap reviewed with your leadership
Weekly active use by team
- Sales
- Support
- Operations
- Legal
- People
This quarter
- Model upgrade, tested on evals firstShipped
- Supplier renewals workflowShipped
- Incidents2 resolved
Next quarter — proposed
- Support triage workflow
- Ticketing connector into the lake
- Board update drafting for Leadership
Start with the problem costing you the most.
It begins with a discovery call, not a proposal. Bring the report, the process or the question that eats your week.
Book a discovery call