AI consulting and workflow automation
Limited time · 40% off an audit. Terms
The audit comes first — nothing else can be priced honestly without it. Everything after is optional. Taking the report to your own developer is a fine outcome.
Where the hours actually go
It is rarely a software problem. It usually looks like this:
- The same information typed into two different systems
- The same question answered from scratch forty times a week
- Two reports reconciled by hand on a Friday, because the systems disagree
Another tool does not fix that — the process underneath has to change.
1. The audit
Two to four weeks. Ends in a ranked list and a working prototype of the best one.
What happens
- Interviews: owner first, then operations, customer-facing staff and finance
- A real sample counted on the slowest workflow, not an estimate
- A systems and data review: what you pay for, who owns it, what is used
- A check for AI features you already pay for that nobody switched on
Managers describe the process as intended; staff describe the real one.
What you get
- A workflow map per process, including the parts nobody documented
- Every opportunity scored on data, workflow, tooling and adoption
- A dollar estimate on each, with a downside case beside it
- A working prototype of the top opportunity, on your own sample data
- A plain statement of what is not ready, and what would make it ready
2. Implementation optional
Two to six weeks per workflow, at a fixed fee once scope is known.
How it is built
- In accounts you own, on tools you already pay for where possible
- Tested against acceptance criteria agreed before work starts
- A human stays in the loop on anything customer-facing
- Models can be swapped, so you are not tied to one vendor's pricing
- Documentation at handover: how to change it, how to turn it off
The kinds of things teams ask us to build
Repetitive inbound
Sorting, drafting and routing the repetitive half, so your team handles the half that needs judgment.
Exception handling
Orders and jobs that fall out of the normal path, caught before the customer notices.
Documents into records
Quotes, orders and invoices arriving as PDFs, then typed in by hand.
Reconciliation between systems
Two systems that disagree, matched automatically with only the real differences surfaced.
Quotes and inbound requests
An incoming request drafted with the right pricing, terms and history filled in.
You already know which costs you most. The audit puts a number on it; what gets built first is your call.
3. Team workshops optional
Role-based training on your own workflows, if your team wants it. Not something we push.
A build only pays off if people use it.
- Scoped by role — customer service needs different things from finance
- Run on your real processes, not a generic overview
- People leave with prompts and procedures that apply on Monday
A note on framing
Every session opens the same way: the goal is removing repetitive work, not people. Staff who fear layoffs withhold how the work really runs.
4. Monthly support optional
Models and suppliers change. Handle it in-house if you prefer; a retainer exists if you would rather not.
What a retainer covers
- Monitoring, so failures are caught before a customer sees them
- Fixes when something breaks or a supplier changes a format
- A monthly review of what the system did and what it cost
- Somewhere to send questions instead of working around a problem
- Build time each month, if you want the roadmap to keep moving
What yours covers is agreed after the audit.
Why we will not sell a retainer first
Without a known backlog there is no honest way to scope ongoing work. A limit we put on ourselves.
Private AI deployment
An add-on for teams whose data genuinely cannot leave their own environment.
“Data must stay in-house” can mean four different things:
- Not used to train models
- Stays in your own cloud account
- Stays in the US
- Never leaves your network
The first three run on managed models inside your own cloud account — the default. Only the fourth needs self-hosted models, which cost the same whether or not anyone uses them. We will tell you which one you need.
Watch the whole data path
A private model does not help if an automation routes the same data through a third-party tool. Every step gets designed and documented, not just the model.
What we do not sell
Saying this out loud saves us both a call.
- Custom machine-learning model training. Different specialism; we refer it out.
- Headcount reduction. We sell time returned, faster response and fewer errors.
- New subscriptions as the deliverable. We use what you already pay for first.
- Retainers before an audit. There would be no defined scope.
- Free audits. A free assessment is a sales funnel wearing a different hat.
Start wherever makes sense
Know where the time goes? Start with an audit. Not sure anything here is worth doing? Start with a $999 Workflow Review.
Limited time · 40% off an audit. Terms