Enterprise AGI Was Achieved in February 2026 | Bernardo Aceituno artwork

Enterprise AGI Was Achieved in February 2026 | Bernardo Aceituno

MTS

October 17, 2026

Stack AI President Bernardo Aceituno details how enterprise AGI was achieved in February 2026, why companies are pivoting en masse toward open-weights models, and how AI agents are struggling against an internet designed solely for humans. Turn ideas into software people love.

0:00

Bernardo Aceituno

I think that February 2026 was the crossing point at which, at least from our own internal buy benchmarks, AGI was achieved for the purpose of enterprise automation. Really, the five things that make an AI agent relevant for the enterprise, which is understanding unstructured data, querying unstructured data, being able to orchestrate tools and actions in a non-terministic way, the ability to interact with visual interfaces, and the ability to synthesize unstructured information, you know, either documents or tools, what not, presentation, what not.

Those five things became really good to a level of reliability that matches human quality by the beginning of this year.

0:38

SPEAKER_2

Hello everyone, and welcome back to MTS. Today, we are joined live in the office here with Bernardo Aceituno, who is the co-founder and president of Stack AI. So Bernardo, welcome to MTS.

0:48

Bernardo Aceituno

Thank you so much. A pleasure to be here.

0:50

SPEAKER_2

I know you're saying you're a first time guest, a long time listener, so you officially got to be here. I think Stack AI has such an interesting story. I know you guys were acquired by Asana earlier this year, but I think one of the big reasons I was excited to talk to you is with your role at Stack AI, you get to see everything that's happening in the enterprise. You get to see it firsthand. Tell me a little bit of the insights and trends that are unique to now 2026 that weren't true a few years ago.

1:17

Bernardo Aceituno

Absolutely. For those who don't know, we at Stack AI, we have a platform that enables enterprises to build custom AGI harnesses for automation of their internal back office processes. We work with the largest law firm in the country, largest head firm in the country, with large companies like BAE Systems, White retirement, we help them come and take AI agents and tailor them for the specific back office functions that they care about, whether if it's commerce, whether if it's compliance, legal, and do so in heavily regulated environments, whether if it's on-premise on aircraft networks or doing so in IBM mainframe machines, we bring AI agents to these environments.

Some of the things that we have seen, of course, progress in the past year has been a few things. I think that February 2026 was the crossing point at which, at least from our own internal buy benchmarks, AGI was achieved for the purpose of enterprise automation. Really like the five things that make an AI agent relevant for the enterprise, which is understanding on structure data, querying on structure data, being able to orchestrate tools and actions in a non-terministic way, the ability to interact with visual interfaces and the ability to synthesize on structure information, you know, documents or tools, what not, presentation, what not.

Those five things became really good to a level of reliability that matches human quality by the beginning of this year. And you could see, I could, I noticed an inflection point as to where our customers are to take problems, as to where the challenge is coming from, where from. I would remember like perhaps two years ago, the biggest challenge was prompting in here. Last year, a lot of it was like the API and NCP server design. Now, a lot of it is kind of control around governance and how they're like, now that you have these systems that work very reliably towards the task, and are quite comprehensive towards it, how do you control what they do? How do you govern what they do? How do you ensure that only the right people get access to it? So we have to make a big effort towards designing a platform that could ensure that, and build an FD team that had that first in mind.

3:11

SPEAKER_2

Interesting. So AGI was fully achieved in February 2026

3:15

Bernardo Aceituno

Enterprise AGI.

3:16

SPEAKER_2

Enterprise AGI. What metric do you have to make that claim?

3:19

Bernardo Aceituno

Vibes. But the truth is that we internally, we have a set of tasks that we care about, that we say, look, and ties back to these five skills I was talking to, where we would say, like, if an LLM can come and receive a task oriented towards this, and achieve it reliably enough, with at least the level of quality of a human, given the correct prom and the correct set of tools, it is for all the intents of purposes of that task, AGI. So when we pick that as a benchmark, we say, how can we achieve these metrics? All the benchmarks out there, where if it's a kind of like computer bench or enterprise bench or whatnot, all of them do kind of like point towards the improvement of this to a level that exceeds human reliability.

Read the other 19 minutes of this episode here.

$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.

Not what you expected? Full refund within 14 days.

Thousands of transcripts fetched by people building searchable podcast archives

Fetch transcripts with code

The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:

request
curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.

Using your own key:

request
curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000796769640