← Blog

Building for practitioners

Suraj Pratap

April 10, 2026

At some point in every technology wave, the people talking about the technology start outnumbering the people using it. AI has been there for a while now. Conferences are packed. LinkedIn is loud. Everyone has a take.

But there's a quieter group that's actually shipping with this stuff — running support operations, processing documents, coordinating teams. AI is part of how they work. One tool among many.

Nedati exists for that group.

Two audiences

The AI industry has quietly split into two camps, and most companies have picked a side without saying so.

There are evangelists — people whose work centres on AI as a subject. They write about it, invest in it, build demos, give talks. They care about novelty and capability ceilings. They want the most impressive thing.

And there are practitioners — people whose work uses AI as a tool. They run it, rely on it, explain its outputs to people who don't care how it works, and deal with it when it breaks. They want the most reliable thing.

These groups want fundamentally different products. One wants to be surprised. The other never does.

What practitioners ask for

Talk to people who use AI in production and you hear the same things:

  • I need to know what it's going to do before it does it.
  • I need to explain what it did to someone who doesn't care about AI.
  • I need it to work the same way on Friday as it did on Monday.
  • I need it to stop when it doesn't know what to do, instead of guessing.

These are trust problems. And you build trust by making things predictable, auditable, and honest about their limits — you don't build it by adding more features.

How we think about products

This shapes every decision we make at Nedati. When we're designing something, we keep coming back to three questions:

01

Is it deterministic?

Same inputs, same outputs. If we can't guarantee that, we make the variance obvious.

02

Is it transparent?

Can you explain what happened to a manager, a compliance officer, or an unhappy customer? If the answer only makes sense to an engineer, we're not done.

03

Does it know when to stop?

When the system hits something outside its boundaries, it should escalate. We'd rather it ask for help than improvise.

Same principle, everywhere

Everything we ship follows the same idea: do one thing, do it well, be upfront about what it does.

Kindflow turns written procedures into AI agents that follow them. When the procedure doesn't cover a situation, the agent stops and escalates instead of winging it.

Extreme Vault manages secrets with zero-knowledge encryption. The architecture means we genuinely can't see your data — that's a cryptographic guarantee, not a promise.

Keep Me Updated sends you a digest of things you told it you care about. You define the topics, it watches for changes, and that's the whole thing.

Stay Awake prevents your Mac from sleeping. One click. No account. No telemetry. It collects zero data. It might be the most honest piece of software we've made.

We'll keep building products this way. The specific tools will change, but the approach won't.

Just work

We don't have strong opinions about AGI timelines or which model will top the benchmarks next quarter.

We care about what makes a product useful at 3pm on a Tuesday, when someone on your team needs it to just work.

That's who we build for. That's Nedati.