Birkan Icacan

Product @RogoAI. Previously @enterpret_ai ('25-'26), @NotionHQ ('22-'25), @Stripe ('21-'22), @Google ('11-'21).

🎓Electronics Engineer & MBA @Sabanciu

[Birkan Icacan @bicacan]

SAML SSO alone won’t win an enterprise deal. But without it, you may never reach the conversation that does.

After seeing this journey at Google, Notion, and Enterpret, I wrote about why going upmarket is a company-wide choice, not a feature/pricing tier. https://t.co/BmS1jaH0eZ

[Birkan Icacan @bicacan]

@iampgzp @NotionHQ Huge congrats Ji and the team!

[Birkan Icacan @bicacan]

Exciting personal news. I joined @RogoAI to lead the Enterprise product team!

I’m bullish on bringing the power of AI into the application layer. As models become more capable, so much of the product opportunity lies in the context and workflows we build around them. Knowledge that AI can access, the tools it can use, and how closely the whole product is tailored to the people and domain it serves.

That’s what Rogo does so well for finance. It brings together financial expertise, a firm’s own knowledge, and agents designed around how bankers and investors work. The opportunity to rethink those workflows across finance is enormous.

At Google, Stripe, and Notion, I helped build products that teams rely on every day and enterprises can trust at scale. At Enterpret, I went deeper into turning fragmented customer feedback into intelligence and action with AI.

Rogo brings those experiences together. The opportunity to help shape how an entire industry works is what makes me so excited about this next chapter.

Thank you to @GabeStengel, @stribwal, and the entire team for your trust and confidence. Every conversation made me more excited about the people, the product, and the opportunity ahead.

And we’re hiring! If you want to help build the AI platform for finance, let’s build together!

[Birkan Icacan @bicacan]

End users fall in love with features. Procurement signs on math.

Your demo wows the team with slick UI, then the deal stalls in late review.

The buyer doesn't care how it works. They care what it saves: support deflection, churn protected, eng hours back.

Sell the math.

[Birkan Icacan @bicacan]

Security compliance like SSO, SCIM, and SOC2 took years to become enterprise table stakes.

AI governance is hitting that exact curve in a fraction of the time.

Teams delaying model auditability and data residency guarantees will get blocked in procurement.

[Max Schoening @mschoening]

make it work
make it good
make it delightful

do not skip steps. demo every day. repeat.

[Birkan Icacan @bicacan]

@yigitkonur @enterpret_ai Thanks for your kindest words Yiğit 🙏 Hope we get to build together one day.

[Birkan Icacan @bicacan]

I recently wrapped up leading Product at @enterpret_ai .

As AI makes it faster and easier to build, the real alpha is shifting toward knowing what to build.

Answers are often hiding in plain sight. Customers tell companies what’s broken, what they need, and what they value through support conversations, sales calls, surveys, reviews, social posts, and product usage. But those signals are fragmented. They rarely reach the right people, influence the roadmap, or shape the decisions that follow.

Enterpret exists to change that.

Over the past year and a half, I’ve been lucky to help Enterpret become a key part of the product development lifecycle by turning fragmented customer feedback into intelligence teams can use to decide what to build, improve the customer experience, and understand the impact of what they ship.

I’m grateful to @varunconfirms and @arnavshr for their trust in me to lead Product. Thank you to my amazing product team (@kaushalvivek, Jay R, @rahulxhq, Sarthak Gupta, Kumar Aniket) and to every xfn partner across design, engineering, GTM, and the broader company. It has been an honor and a great pleasure to build alongside you.

My chapter at Enterpret has ended, but my conviction in the problem and my admiration for the Enterpret team and product are stronger than ever.

I’m rooting hard for Enterpret and the team building its future. More on my next chapter soon.

[Birkan Icacan @bicacan]

Teams reprioritize entire quarters because one enterprise customer escalated a minor feature request loudly.

Meanwhile, the accounts that quietly churned were full of signals no one acted on.

The loudest complaint gets attention. The silent cancellation teaches you more.

[Birkan Icacan @bicacan]

@Johnsjawn Congrats to you both! 🎉

[Birkan Icacan @bicacan]

Nobody in procurement ever bought a product because the demo was impressive. I learned this the hard way across Google Docs and Notion.

At Google, real-time collaboration in Docs was exciting to show. At Notion, modular building blocks were genuinely compelling. But exciting the user and convincing the economic buyer are two different conversations.

Feature selling explains what the product does. Outcome selling shows which measurable business result should change and by how much. That second conversation is the one where the budget actually gets approved.

[Birkan Icacan @bicacan]

At Notion, we worked through SSO, SCIM, audit logs, compliance, admin controls. Every one of those removed an objection. None of them produced a deal.

What actually happened: once security was satisfied, the conversation moved to the economic buyer. And the question was always the same. What is the return on this?

I saw the same pattern again at Enterpret. Security can stop a deal. It rarely wins one. The features that pass the evaluation are different from the features that close it.

[Birkan Icacan @bicacan]

"Fortune 500" describes company size. It does not describe a customer problem.

I have seen companies close a few large deals and declare they have product-market fit in the enterprise. But a large contract is evidence of demand. Repeatability is evidence of a market. Those are different things. Building an entire motion around one logo is how you confuse them.

[Birkan Icacan @bicacan]

One AI agent may support hundreds of employees. How do you charge per seat for that?

Seat-based pricing worked for enterprise SaaS because it was legible. A buyer could estimate users, negotiate a contract, plan a budget. AI agents break that model because charging per user disconnects from both usage and value.

Tokens may drive your costs. They are rarely what the customer believes they are buying.

[Birkan Icacan @bicacan]

Google Sheets had to work with Salesforce and SAP. Enterpret has to push tickets into Linear and Jira automatically. Same lesson at both companies.

The enterprise workflow is never "open your product." It is pull data from somewhere, do something with it, push the result somewhere else. If your product is not embedded in that chain, people have to remember to visit it. And they won't.

Okta puts the average app count at companies with 2,000+ employees at 259. You are one of 259.

[Birkan Icacan @bicacan]

Fair skepticism. But I don’t think the opportunity is to replace real research with simulated data. It’s to help teams learn faster before investing in a full study. For many day-to-day product decisions, the choice isn’t simulated data versus real data. It’s simulated data versus no data at all.

In the AI era, time to market will increasingly separate great companies from good ones. Coding agents are collapsing the execution stage. Teams are building software factories and shipping in days. But learning and validation haven’t accelerated at the same rate.

The risk is obvious: we’re getting much better at building the wrong things faster.

Let's look at the actual options:

1. Paid ads can be useful for quick validation, but they don’t work for every audience. If your ICP is a CXO at a Fortune 5000 company, you may not be able to reach or learn much from them through ads. Some audiences simply aren’t buyable.

2. Traditional surveys aren’t exactly working beautifully either. Have you used Qualtrics, or a similar tool, lately? Not the most intuitive tool out there, and even once you ship the survey, fill rates run 2-3%. People won't answer a 1-5 rating question, let alone an open-ended one. Insight quality is eroding fast.

3. Then there’s user research. Nothing beats talking to users directly, and I don't think simulation replaces that. But a good study takes ~3-4 weeks (recruiting the right people, writing the brief and the interview script, running the sessions, synthesizing, producing the report ) and it's expensive before you even count the incentives (e.g. gift cards) you need to get decent participation. Listen Labs and others are compressing the cycle, but a single study there still isn't cheap. So you can't run one for every hypothesis, feature, or solution a team is weighing. Most UXR teams already have a backlog. I've been the PM waiting for a research request to get picked up... and the feature shipped anyway, on intuition.

So the question isn’t, “Can Simile predict human behavior perfectly?” Probably not.

It is “Can it give a founder or PM a directionally reliable answer from their target ICP in minutes, helping them kill bad ideas earlier and identify where deeper research is worth the time and money?”

If it can do that consistently, it would be a game changer.

[Birkan Icacan @bicacan]

Hardest part of using AI today isn't the model. It's knowing what to ask it.

ChatGPT and Claude are remarkable. But they're still largely reactive (at least for now). They wait for us to frame the right question. And most of us aren't good at that. We don't always know what we want, let alone how to phrase it.

What's even worse is by the time you figure out what you should have asked, the moment to act has usually passed.

Real promise of AI agents is not simply better answers on demand. It’s better and timely signal detection.

An agent that notices the customer risk, the emerging opportunity, the slipping commitment, or the decision you’re avoiding and brings it to you before you think to ask changes the nature of the tool.

That is when AI becomes less like a search box and more like a capable colleague quietly paying attention, connecting dots, and helping at the moment it matters.

The future of AI is not just conversational.

It’s ambient, proactive, and trusted enough to be useful before the question exists.

[Birkan Icacan @bicacan]

@ericzakariasson Great workflow Eric. You may also consider adding one more step to this workflow and automate closing the loop with the users who reported the feedback that you've addressed their feedback 🙂

[Birkan Icacan @bicacan]

@eoflazer @NotionHQ @humansand Glad that our paths crossed. All the best in your next adventure!

[Birkan Icacan @bicacan]

Customer feedback has outgrown dashboards and chat threads. Imagine this workflow:

A "Quality Monitor" Agent detects a spike in customer feedback.

It pulls supporting evidence from @enterpret_ai, checks product analytics in @posthog, finds the likely culprit PR in @github, creates a @Jira or @linear ticket with the full issue summary, and alerts the right team in @SlackHQ.

Then it keeps going.

When the ticket is resolved, the agent identifies the impacted users, drafts personalized customer follow-up messages, and notifies the team so they can close the loop.

Three days after the fix ships, it runs a post-launch impact analysis to show whether the issue actually improved.

That is the shift we launched yesterday with Agent OS at Enterpret.

Not just a dashboard.
Not just a chat assistant.

An operating system for customer intelligence agents.
Agent OS brings together:

1. Enterpret Agent: Your Customer Intelligence Agent (no, not that CIA) that you can collaborate with in real time.

2. Agent Automations: Recurring and signal-driven workflows that detect, investigate, act, and follow through.

Quality monitoring is just one example.

You can apply the same pattern to weekly VoC reports, account health monitors, churn risk detection, help center gap analysis, launch monitoring, win/loss analysis, and more.

Customer intelligence should not stop at “what did customers say?”

It should help teams understand what changed, why it matters, who is affected, what to do next, and whether the fix worked.

That’s what we’re building with Agent OS.

Proud of the Enterpret team 🚀