Summary
You’ll have one workflow running on agreed rules, with a clear “needs a human” path, plus a dashboard that shows what ran, what failed, and what changed in your key systems.
You reached out because a messy month showed how easy it is for things to slip. We build with exception rules, sign-offs, and alerts, so mistakes don’t hide until month-end. Next, we’ll map one workflow and agree the pass and fail cases.
We start by writing down the workflow as it really happens, including the odd cases that blow things up at month-end. Then we check the data, permissions, and handoffs in each system before we automate anything. We ship a pilot you can watch, fix what it misses, then lock in monitoring and a runbook.
Who you are working with
We’re a small Data and AI team that got tired of automations that only work in a demo. Tom started in operations automation, then moved into AI once we could make classification and routing reliable enough to trust with billing and customer comms. We build around the systems small teams already use, and we leave you with a runbook you can follow.
We get picked when the risk is not “will it save time” but “what happens when it’s wrong”. We don’t ship anything that can write back into Xero or a CRM without a logged decision path, rollback options, and a place to review exceptions. You’ll see the rules, the edge cases, and the alerts before the workflow touches live data.
The team
You’ll deal with the same four people from first workshop through to monitoring, so decisions don’t get lost between calls.
Team
Tom Baxter, Automation lead. Tom leads the automation, runs the workshop, and signs off the exception rules before anything goes into a live system.
Holly McLeod, Data and systems audit. Holly checks your data and system permissions, so the automation reads the right fields and only writes where you’ve agreed it can.
Ethan Caldwell, AI workflow engineer. Ethan builds the AI workflow and the integrations, and keeps the logic readable so you can change it later without guesswork.
Priya Nair, Testing and monitoring. Priya tests with real edge cases, sets up the monitoring and alerts, and makes sure you can see failures before they become month-end surprises.
Recent projects
Here are three recent AI workflow automation builds where the win was fewer slips, faster follow-up, and a clear exception queue.
Portfolio
Quote follow-up automation. Mapped the quoting flow, then built an inbox-to-CRM capture with an exception queue for ambiguous jobs. Follow-ups went out the same day instead of waiting for an admin gap.
Invoice and job sync. Connected job completion notes to invoice drafts, with checks for missing PO numbers and mismatched rates. The workflow stopped incomplete invoices before they hit the ledger.
Stock update routing. Automated stock adjustments from sales channels into a single spreadsheet-backed master, with warnings when SKUs didn’t match or counts went negative. Updates stopped bouncing between inboxes.
“Nothing runs silently now. If it’s unsure, it flags it fast.”
Owner-manager, a small trade business
Discovery workshop notes
A plain-English map of the workflow you choose, written as steps, inputs, outputs, and owners. This arrives after the workshop as a doc you can sign off.
EXCEPTION AND SIGN-OFF LIST
A list of the cases where the automation must stop, ask, or route for review, plus what “good” looks like. This is agreed before we connect live systems.
SYSTEMS AND DATA AUDIT REPORT
What we found in your tools, data quality, and permissions, with a short readiness checklist. This arrives before build so fixes happen once, not mid-flight.
Automation build in your tools
The working AI workflow automation, connected to your agreed systems, with the rules and prompts documented. This is delivered into your account, not ours, at handover.
RUNBOOK AND HANDOVER PACK
A step-by-step “what to do when” guide: what to check daily, what to do on a failure, and how to change common rules safely. This arrives at handover.
MONITORING DASHBOARD AND ALERTS
A dashboard that shows runs, failures, and held items, plus alert rules that notify the right person. This runs every month, with tweaks as your workflow changes.
What it costs
Most teams start with one workflow, prove it in a pilot, then decide if they want us to keep monitoring it month to month. Prices are per month.
Priced items
Next steps
If you’re ready to stop month-end surprises, the next step is to lock in the first workflow and the access we’ll need to build it safely.
1. Sign this proposal so we can book the workshop. 2. Pay 50% of the setup invoice to start. 3. Join the workshop and share system access for the audit.
WHAT SYSTEMS ARE YOU GOING TO CONNECT TO, AND WHAT ACCESS DO YOU NEED?
We only connect what the chosen workflow needs. We’ll ask for read access first, then add write access only for the specific actions you approve. Where we can, we use service accounts and scoped permissions so access stays tight.
WHAT HAPPENS WHEN THE AI GETS IT WRONG OR CAN’T CLASSIFY SOMETHING?
It does not guess and carry on. We build a “held for review” path with reasons, a queue, and alerting, so a person can decide. We also log the misses so we can tune prompts, rules, or training on a schedule.
CAN WE START WITH ONE WORKFLOW WITHOUT RIPPING UP EVERYTHING?
Yes. We usually start with one workflow that has a clear start, a clear finish, and a place where admin time is being burned. We build it so it can sit alongside your current process while the pilot runs.
WHO MAINTAINS IT WHEN SOMEONE CHANGES A FORM OR A PROCESS?
You’ll have a runbook for small changes, plus a clear “change hits this part” map. If you want us to maintain it, we’ll monitor the workflow and update rules when your forms, fields, or handoffs change.
Signature
Fee summary
Payment
Setup work is paid 50% to start and 50% when we hand over the runbook and monitoring dashboard. Ongoing monitoring is invoiced at the start of each month. All invoices are payable within 7 days.
Scope changes. We’ll work from the signed-off exception list and workflow steps. If you want to add steps, new systems, or extra outputs, we’ll write the change down with the price and timing before we build it.
Access. We only request access the workflow needs. We start with read access, then add write access only for the actions you approve. Where we can, we use service accounts and scoped permissions.
Your turnaround. We’ll need a main contact to answer questions and sign off exceptions, plus timely access to the systems involved. If we’re waiting on access or sign-off, the timeline pauses until we have what we need.
Held for review
When the AI can’t classify something or confidence is low, the workflow routes it to “held for review” with a reason, a queue, and alerts. A person makes the call before anything gets posted or sent.
Pilot period. We’ll run the workflow alongside your current process during the pilot so you can compare results. We’ll log misses and edge cases during the pilot and apply fixes before you rely on it day to day.
Ownership. Once the final setup payment is made, you own the automation build we deliver in your tools, plus the runbook and handover pack. Any third-party platforms stay under their own licence terms.
Confidentiality. We keep your customer, job, and financial data confidential and only use it to build and support the agreed workflow. We’ll share access only with our team members working on this project.
