Launching Kiia: AI agents for everyday work
I want to share a project of our own: we are launching Kiia, our agency building AI agents for businesses.
If you read this blog, you know my interest in building software and bringing AI into development. I have shared how I work with coding agents, including the benefits and the difficulties of coordinating them. With Kiia, I want to bring that conversation to the work happening in other parts of a business: which task needs help, what information an agent should access, and how we check that it did the right thing.
I am excited to give that work a home of its own and introduce it here.
What we want to build
At Kiia, we create agents that connect to the tools a company already uses and carry out tasks within a defined process. The starting point might be a CRM, a database, or internal documentation. What matters is understanding the result the person on the other end needs.
Imagine a sales report that requires reviewing CRM opportunities, matching them against payments, and preparing a summary for the sales manager. An agent could gather that information and deliver the report through the channel the team uses. First, we would need to agree on which sources to query, how to handle incomplete data, and who reviews discrepancies.
That is an illustrative example. Every implementation depends on the available access and the company's rules. Two tools having an API still leaves plenty of work to figure out.
Starting with one specific process
My preference is to start with a frequent task whose output we can review. That lets us compare the work before and after the pilot, understand exceptions, and decide whether to expand the scope.
It also means choosing the solution around the problem. A systems integration or a fixed-rule automation might be enough. When the task requires interpreting information or working with documents, we assess where AI makes sense.
Our delivery methodology explains how we map the process, verify permissions, and validate a pilot that can be stopped. We keep any constraints that still need checking visible. I would rather resolve those questions early than discover them when the team already depends on the workflow.
The people who will use it
One part I care about is what happens after handoff. Someone needs to know how to review the agent's work, handle an exception, and pause the process if something goes wrong.
That is why the scope also needs to include ownership and documentation. Sensitive steps can require human approval; the level of autonomy is agreed with the team. The tool needs to fit that way of working and respect data access boundaries.
An invitation
If you have a process that currently relies on moving information between tools by hand, I would like to hear about it. On Kiia's website, you can book a conversation to review what could be connected and what we would need to verify first.
And if you know someone considering AI agents for their business, sharing this launch with them is a way to help us introduce the agency.