HatchWorks

HatchWorks We build AI-native solutions — and use AI to build software better, faster, smarter.

We made the list! 🎉HatchWorks AI is part of Inc. Magazine's 2026 Awards in the AI, Technology & Development category. ...
10/20/2026

We made the list! 🎉

HatchWorks AI is part of Inc. Magazine's 2026 Awards in the AI, Technology & Development category. The awards recognize top B2B companies across the globe.

Power Partners are chosen from client feedback. Inc. asks the people we work with how likely they are to recommend us. This is our third time on the list, and we owe it to our clients. Thank you.

👇 Link in the comments

10/20/2026

In two years, renting your most important IP is going to look strange.

That is Manos Koukoumidis's answer on the latest Talking AI clip, when Matt Paige asks what will feel completely normal by then that sounds odd today.

The cofounder and CEO of Oumi expects leaders to look back and wonder how they ever thought it was fine to rent their core intelligence, or to outsource its development to another organization.

Right now that is the default. Plenty of companies are still renting large models for the work that defines them.

Manos thinks the opposite will read as obvious: build it, own it, and keep improving it, the way you would with any other asset that decides whether you win.

Every default looks permanent until it does not.

👇 Link in the comments

In September, OpenAI passed Anthropic in OpenRouter wallet share for the first time since February 2024.If your AI stack...
10/20/2026

In September, OpenAI passed Anthropic in OpenRouter wallet share for the first time since February 2024.

If your AI stack is tuned to one model, that kind of swing gets expensive. Skills, tool connections, memory, and guardrails all get built around one model's behavior. Swapping the API is the easy part.

The fix is a neutral AI harness:
- Routing that sends each task to the right model
- Tool and data connections built once, using protocols like MCP
- Eval suites that run the same tasks against any new model
- Guardrails and memory that don't depend on one provider

Build the harness once, and the model becomes a part you can swap.

Coinbase says it cut its AI bill nearly in half while token usage kept climbing.How? It wasn't one cheaper model. It was...
10/20/2026

Coinbase says it cut its AI bill nearly in half while token usage kept climbing.

How? It wasn't one cheaper model. It was the system around a portfolio of models:
- Cheaper defaults for routine work
- Task-aware routing through an internal gateway
- Prompt caching, with one implementation going from about 5% to 60% cache hits
- Leaner context
- Visible usage tracking

Meanwhile, pull requests per engineer rose 2.2x year over year.

The broader point: give each workload only the intelligence, context, control, and verification it needs.

👇 Full breakdown below

10/19/2026

Omar Shanti is an early bird. His AI agents are night owls.

Mariana takes our CTO through a rapid-fire round in Bogotá, at the end of two weeks visiting our San José, Medellín, and Bogotá offices.

What we learned:
☕ Coffee over tea, and only from Andrea Portillo’s farm in Costa Rica
🚀 Deploy on a Friday? Yes. With solid CI/CD, deployments shouldn’t scare you
🖊️ Huddle or Slack? Whiteboard first, so huddle wins

Asked to finish the sentence “You know you work at HatchWorks AI when...”, Omar said every day is different. What stays the same is a culture of helping each other and servant leadership. For him, that means being part of everything from delivery to hiring to mentorship to the internship program.

Which side are you on: Friday deploys, yes or never?

👇 Link to our careers page in the comments.

10/16/2026

What does it actually mean to work at an AI-native company?

Mariana asks our CTO Omar Shanti that question. His answer: AI isn’t only something we build for clients. It’s how we run the business. Every task starts with one check: can AI help us do this better and faster? Whatever time that saves goes to the work that sets us apart.

Omar also shares the bet he thinks other AI companies will copy soon. HatchWorks AI is going all in on the Forward Deployed Engineer role. He sees it as the next step for consultants who already know the craft, think like the business, and build real client relationships.

His message to anyone thinking about joining: lots of room to grow, real leadership opportunities, and a great ride.

👇 Explore open roles at HatchWorks AI. Link in the comments.

Picture a COO at a general contractor spending a Tuesday in a hotel conference room with twelve people and one facilitat...
10/16/2026

Picture a COO at a general contractor spending a Tuesday in a hotel conference room with twelve people and one facilitator. Two days, 15 pre-interviews, dot voting, an impact-versus-effort grid.

By Thursday, there's a list of AI ideas. Procurement, Safety, and HR never had anyone in the room.

There are three things that format can't see:
- Coverage: missing functions look like functions with no opportunities
- Voice: the most senior, confident people set the agenda
- Imagination: people name only the AI they've already seen

AI-guided interviews remove the cost limit on how many people you can hear from. Strategists still make the calls.

👇 Full article below

Please welcome Mona Champaneri, our newest Sr. Client Partner at HatchWorks AI! 🎉Mona will partner with business and tec...
10/15/2026

Please welcome Mona Champaneri, our newest Sr. Client Partner at HatchWorks AI! 🎉

Mona will partner with business and technology leaders to turn AI ambitions into working software, from strategy through delivery. If you're exploring what AI could do for your team, she's a great person to know.

Welcome to the team, Mona!

AI pilots rarely die because of the model. Here's where they actually stall:1. Data that looked ready wasn't2. Integrati...
10/15/2026

AI pilots rarely die because of the model. Here's where they actually stall:

1. Data that looked ready wasn't
2. Integration nobody scoped
3. Governance and compliance showed up in month four
4. The workflow only existed in someone's head
5. Nobody owned the gap to production

Flip each one into a requirement and you get a job description: someone embedded, accountable, and responsible for production. That's a forward deployed engineer.

The rules for a pilot that is allowed to succeed: pick for value and buildability, assess data before promising dates, invite governance in on day one, and name one owner all the way to production.

👇 Full article below

A few months after ChatGPT launched, one of the very first PaLM customers asked Manos Koukoumidis for the model weights....
10/14/2026

A few months after ChatGPT launched, one of the very first PaLM customers asked Manos Koukoumidis for the model weights.

He had to say no. The terms of use did not allow it.

So they asked if they could distill PaLM into their own internal version. Also no.

Manos was running the natural language services at Google Cloud at the time, and the request stuck with him. Why would a major organization push that hard for a model that was complex to build, complex to maintain, and expensive to run? By early 2023 he had his answer. They would not be the last to ask.

On the latest Talking AI episode, Manos, now cofounder and CEO of Oumi, explains why he left building Gemini to build pretty much the opposite: a platform that helps any enterprise create, deploy, and keep improving models it owns.

His rule is short. Rent what is generic. Build and own what differentiates you, or what costs you the most.

The customer asking for the weights was early. It was not wrong.

👉 Link in the comments

Address

5256 Peachtree Road NE Ste. #140
Atlanta, GA
30341

Opening Hours

Monday 8am - 6pm
Tuesday 8am - 6pm
Wednesday 8am - 6pm
Thursday 8am - 6pm
Friday 8am - 6pm

Website

Alerts

Be the first to know and let us send you an email when HatchWorks posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to HatchWorks:

Shortcuts

Share

Category