“Hey, can you help me?” The phrase that breaks traditional automation
Forms and workflows only work when people fill them in. Most requests arrive as a sentence in a chat. That is where automation has always failed, and where AI Workers start.
28 Aug 2026 · 2 min read

Every operations team has a portal. It has forms for access requests, equipment, leave, expenses, and a dozen other things. And every operations team knows that most requests never touch it. They arrive as a message: “Hey, can you help me with this?”
Why forms lose
A form asks the person to translate their problem into the organisation's categories before anyone will listen. A message lets them describe the problem in their own words and hand it to someone who understands it. Given the choice, people pick the message every time, and no amount of training changes that.
Traditional automation was built on the form. Workflow tools, ticket routing, robotic process automation: all of it assumes structured input. The moment the input is a sentence, a human has to read it, classify it, and re-key it. The cost of that step is why so much automation delivered less than promised.
What changes with an AI Worker
An AI Worker starts from the sentence. It reads the request, works out what is being asked, checks the policy that applies, and either acts or asks a clarifying question. The form still exists, but the worker fills it in, not the person.
- “I can't push to the repo” becomes an access check against the policy, then a group change or an escalation.
- “New starter on Monday, same as Ama” becomes an onboarding run cloned from an existing profile, with approvals where the policy requires them.
- “Why was I charged twice?” becomes a transaction lookup, a reconciliation, and a refund request routed to the right approver.
Meet people where they already are
This is why Kepler lives in Slack and Teams rather than in yet another portal. The channel your staff already use becomes the front door, the worker handles the translation, and your ITSM stays the system of record. Nobody has to learn anything new for the automation to work.
The most expensive phrase in operations is not a complaint. It is a request that a person had to re-type.
More from the blog

Market5 Sept 20262 min read
Nine enterprise AI predictions for 2027
AI Workers replace chatbots, pilots get shorter, and the data moat becomes the only moat. What we expect to see across African enterprises next year.
Read
IT operations12 Aug 20262 min read
Automated ticketing: how AI Workers cut IT workloads and speed up SLA resolution
Up to four in ten IT tickets are repetitive enough to be resolved end to end without a human. Here is where the time goes today and how an AI Worker gets it back.
Read