We started BAKD AI because we kept watching small teams lose their best hours to work that should never have been theirs in the first place. A founder copying an order from WhatsApp into a spreadsheet. A support lead retyping the same refund policy for the tenth time that week. An ops manager hunting through a Notion page that nobody has updated since last year.
The work itself is not hard. The problem is that it is spread across a dozen places, none of which talk to each other.
The real problem is not a missing chatbot
When people hear “AI for SMEs,” they often picture a chatbot bolted onto a website. That is not the problem we are trying to solve. A chatbot can answer a question, but it cannot draft a reply in the voice of your business, check stock in your spreadsheet, log the conversation in your CRM, or escalate to a human when something looks off.
Real work involves judgement, context and follow-through. Generic chatbots fall short because they treat every question as a one-shot Q&A. There is no memory of your business, no grounding in your documents, no record of what was decided and no way for you to step in before something goes out the door.
What we mean by AI employees
When we say “AI employee,” we mean something more specific than a model behind a chat box. An AI employee, in our view, has:
- A job description. A clear scope, a tone of voice, and a set of tools it is allowed to use.
- A knowledge base. Approved documents, policies and product info it can quote from, with citations.
- A workspace. The systems it can read from and act in, scoped to least privilege.
- A manager. A human who reviews drafts, approves anything that leaves the building, and can see every step in an audit log.
Frame it that way and the design problem changes. We are not building a smarter autocomplete. We are building a way to hire, train and supervise software that does real work on behalf of a small team.
Why SMEs, why now
Most of the AI tooling we see today is built for engineering teams at large companies. It assumes you have a platform team, an evals team and a budget to integrate models into bespoke pipelines. That is fine for the top of the market. It does very little for the cafe owner running operations from a phone, the agency juggling six client inboxes, or the clinic chasing follow-ups across two spreadsheets.
Foundation models are now good enough that a small team should be able to set up an AI employee in an afternoon, without hiring an AI engineer. The missing piece is not the model. It is the product around it: the knowledge base, the approvals, the audit log, the integrations, and a UI that respects the way SMEs actually work.
What we are building first
BAKD AI is at the prototype stage. The first slice is intentionally narrow: a workspace where a team can upload their knowledge, define one AI employee, point it at a single channel, and have a human approve every reply before it goes out. That is enough to learn whether the shape of the product matches the shape of the problem.
From there we plan to grow into operations and sales, connect more tools, and gradually let the AI act on its own where the risk is low and the audit trail is good. We will publish what we learn as we go.
The honest part
We do not have shipped customers. We do not have funding announcements to make. We do not have a finished product to demo. What we have is a small team in Kuala Lumpur, a clear view of the problem, a working prototype, and a list of businesses willing to try it with us.
If that sounds like your team, we would love to talk.