How to Make a Company AI-Native: Michael Batko on the Three Systems to Build First
Making a company AI-native starts with three systems, according to Michael Batko: a shared company AI brain that holds context, a skills library so one person's workflow becomes everyone's, and a command center that shows what's running and how well. Michael spent eight years as CEO of Startmate, often called the Y Combinator of Australia, before founding Hourglass AI, which builds these systems inside other businesses. He explains why he kept it a service business, which companies make the best clients, why behavior change is harder than the technology, and how he prices audits, builds and ongoing support.
Guest
Michael Batko
Founder, Hourglass AI
Michael Batko is the founder of Hourglass AI, which audits companies, builds AI systems for them and maintains those systems over time. Before that he spent eight to nine years as CEO of Startmate, often described as the Y Combinator of Australia, backing hundreds of startups. His earlier career included American Express, PwC and KPMG, followed by two startups, one of which went public and one of which was sold.
Key Takeaways
- 1.Customer conversations come first; at Hourglass, clients pay for an audit so they share their real problems.
- 2.Productize internally and sell as a service because clients want one trusted owner.
- 3.The best-fit client is 20 to 100 people, founder-led and without an engineering team.
- 4.Build three foundations first: a company AI brain, a shared skills library and a command center.
- 5.Behavior change is harder than the technology, and clients need to arrive ready for change.
- 6.Use a low flat-fee audit, a scoped build range and ongoing support; clients should own their AI bill.
Full Transcript
The number one thing every founder should focus on is customers and customer conversations. At Hourglass we literally get paid to talk to customers, because we audit the company first.
We build custom two or three times, productize it internally so we can deploy it easily, and still offer it as a service. They want one head on the chopping block.
Our ideal client is 20 to 100 people with no engineering team. Usually founder-run, with a strong leadership team. Get the founder on board and you can transform the whole company within six months.
The first thing we do for every client is create a company AI brain. It's ring-fenced so not everyone sees salary data, but it already has the client data and company tone of voice.
A shared skills platform means one person builds a due diligence skill and the whole firm has it, as one global version.
The hardest thing about AI transformation isn't the technology. It's behavior change. People do not want to change.
A client call finishes, the transcript goes into our AI brain, and it drafts a follow-up with next steps. Within five minutes both people have an email from me.
I hate persuading people to use AI. You can't help someone who doesn't want it. They have to come ready for the change.
We've learned to give a range rather than a fixed fee, then run a scoping call before the build and reprice based on the features.
We never own a client's AI bill. If we take it, nobody at the client cares how much they're spending, and you want that feedback loop.
Frequently Asked Questions
Related Resources
More Episodes
How Do Agencies Prove Their Value When Clients Have AI? Gary Magnone on Outcomes, Pricing and Burnout40 min
What Should an Agency Owner Delegate First? Karl Sakas on Stepping Back, Selling and Getting Paid What Your Agency Is Worth40 min
Building a Repeatable Creative Studio: How Birdhill Delivers Monthly CPG Photography at Scale35 min