
AI workforce
Agents that do the work, not just answer questions
Six agent functions, each wired into your systems with defined permissions and a written record of every action. A chat window waits to be asked; these run when something happens.
The workforce
01
Front desk agent
Answers every enquiry within minutes by phone, form or message, qualifies the job and books against real availability.
02
Quoting agent
Prices from live catalogue and stock, builds a draft order, and never sends anything a human has not approved.
03
Follow up agent
Chases open quotes, abandoned carts and unpaid invoices on a schedule, and stops when the customer responds.
04
Operations agent
Runs the reporting, reconciles records between systems, and flags exceptions with the reason attached.
05
Support agent
Handles repeat questions from your own approved wording and escalates anything unusual with context.
06
Engineering agent
Deploys, monitors and fixes the systems underneath, with every action logged and reversible.

The same work, before and after.
Where the line sits
What the agents may do alone
- Answer, qualify and book within set rules
- Draft quotes and orders for approval
- Chase and remind on schedule
- Report, reconcile and flag exceptions
What always needs a human
- Sending money, refunds and pricing exceptions
- Anything that commits the business contractually
- Unusual situations and complaints
- Deleting records or changing access

How the work moves once the system is in place.
Why this is not just ChatGPT with a prompt
An agent needs a trigger, a permission model, access to your systems of record, a log of what it did and a human approval step for anything that moves money or contacts a customer. That plumbing is the product. The language model is one component inside it, and where a deterministic rule is more reliable we use the rule.
Next Step
Find out which agent pays for itself first.
Twenty minutes, no obligation, and we will tell you what not to automate as well.
