How we work
How an AI automation engagement works
Every engagement follows the same five steps, from a first conversation to a system your team owns and supervises. The pace depends on your business and your systems, and each step ends with a decision you make.
1. Scoping call
The first conversation is about your business, not our tools. We ask what work takes up the most time each week, where enquiries or tasks get lost, which systems you already use, and who on your team is involved. We also ask what should never be handed to software, so those limits are clear from the start.
You come away with a plain summary of what we heard, the areas that look suitable for automation, the ones that do not, and a written proposal for the next step if it makes sense to continue.
What we need from you
- The person who knows the day-to-day process, not only the decision maker
- A list of the software you use for calls, email, scheduling and customer records
- One or two recent examples of the task you want to automate
2. Process review
Next we map the work as it really happens. We walk through each step with the people who do it: where information arrives, who touches it, what decisions are made and where it ends up. Most processes contain a mix of routine steps and judgement calls.
The goal is to separate the two. Routine, repeatable steps are candidates for automation. Judgement calls, exceptions and anything sensitive stay with a person, or get a clear approval step. If a process is not repeatable enough to automate well, we say so.
What we need from you
- Time with the staff who carry out the process
- Access to sample records, with personal details removed where possible
- Your rules for exceptions, refunds, complaints and anything you treat as sensitive
3. Build
We assemble the automation using your existing systems wherever we can, connecting them rather than replacing them. Each automation follows the map agreed in the process review, so every step can be traced back to a decision you made.
Human approval steps are placed where a mistake would matter: before a message goes to a customer, before a record is changed, or before anything involving money or personal information. Data handling rules are set here too: what the automation can read, what it stores, where it stores it and who can see it. Any third-party tools are reviewed before they are connected.
What we need from you
- Admin access, or a named person who can grant access, to the systems involved
- A decision on who approves which actions
- Your existing privacy, retention or security requirements
4. Test with you
Before anything reaches a customer, we run the automation against real cases from your business in a safe setting. Your team checks the results, tries unusual situations and tells us where the output is wrong or unclear.
We adjust and test again until your team is comfortable. Going live is your decision, and it can happen in stages, starting with a small part of the work while people keep an eye on results.
What we need from you
- A set of real past cases, including a few awkward ones
- Staff time to review results and give honest feedback
- A named person to sign off before going live
5. Handover and support
You own the workflows, the configuration and the data in your accounts. We hand over documentation that explains what each automation does, where the approval steps sit, how data moves and what to do if something stops working.
Your team gets a walkthrough on how to operate and supervise the system. After handover, changes are requested, agreed in writing and tested before they go live, so the system stays understood rather than drifting over time.
What we need from you
- The people who will operate the system day to day
- A contact for change requests
- A place to keep the documentation your team can reach
Governance is designed in at step 3, not added afterwards
Human approval, data handling rules and vendor review are part of the build itself. Deciding them early means the automation is shaped around them, instead of having controls bolted on once it is already running. It also means you can explain to staff, customers or an auditor what the system does and who checks it.
The same team also carries out IT audits, cybersecurity assessments and AI governance work, so the questions an auditor would ask are asked while the system is being built.