Why we raised $13.5M to define Process Intelligence for the Agentic Era

Today we're announcing that Tekst has raised $13.5 million in a Series A round led by Elephant, the Boston-based venture firm with a track record of backing category-defining software companies. This round isn't really about the money. It's about what comes next: for Tekst, and for the way enterprises will run themselves over the next decade.

We believe that the next great enterprise software category is being created right now. We call it Process Intelligence for the Agentic era. And we intend to lead it.

The promise of agentic AI is real. The reality is not.

Every CEO we talk to is under pressure to "do something with AI". Every CIO has a stack of pilots. Every COO has been promised that agents will soon run the back office.

And yet, almost no one is actually getting there.

The numbers tell the story. Most enterprise AI projects never make it out of pilot. The ones that do typically automate a thin slice — a single email type, a single ticket category, a single document — and stall there. Agents that demo beautifully on a clean problem fall apart the moment they hit the messy reality of how work actually moves through a company.

This isn't an AI problem. The models are extraordinary. The reasoning is there. The infrastructure is there.

The problem is that AI doesn't understand the business.

It understands data. It doesn't understand the process: the unwritten rules, the exceptions, the handoffs, the customer-specific quirks, the "we always do it this way for this account" knowledge that lives in people's heads and email threads. None of that is in the system of record. None of it is in the LLM's training data. And without it, an agent is, at best, a very expensive assistant.

What's actually missing: the process layer

Look at how a quote-to-cash process really runs in a Fortune 500 company. The CRM shows opportunities. The ERP shows orders. The contract repository shows clauses. But the actual work — how a quote becomes an order, how an exception gets escalated, how a non-standard request gets approved, how a complaint gets routed to the right person — happens between the systems. In inboxes, attachments, Slack threads, side conversations, and the experience of the person who's been doing this for fifteen years.

That layer has been invisible to software for forty years. Process mining tried to surface it, but only saw what was already in the logs of structured systems. RPA tried to automate it, but could only follow brittle, pre-mapped paths. BPM tried to model it, but required armies of consultants and froze the moment reality changed.

What enterprises actually need is a system that observes how work really happens, builds a living model of the process, and then lets AI agents operate inside that model safely and reliably.

That's Process Intelligence. And that's what we've built.

Process Intelligence, defined

Process Intelligence is the layer between an enterprise's systems and its AI agents. It does three things:

It discovers the real process automatically — not by interviewing people for six months, but by observing the digital trail that work already leaves behind: emails, documents, system events, decisions, exceptions. It reconstructs how work actually flows, end-to-end, across systems and people.

It understands the context behind every decision — which customer gets which treatment, which exception goes to which team, which approval is required when, what "done" looks like for this specific business. The unwritten rules become explicit.

It enables AI agents to act inside that understanding. Agents don't operate on raw data and hope for the best. They operate on a living model of the process, with clear boundaries on what they can decide autonomously, when a human is needed, and which exceptions are allowed to pass.

The result: AI agents that are not impressive demos, but trusted colleagues. Agents that handle real volume on real processes, at real enterprises, with the reliability the back office demands.

What this looks like in production

At Becton Dickinson, a global medtech leader and a Tekst customer, our platform was handling 90 percent of the work autonomously within three weeks — on a process the company had been trying to crack internally for years. At Nokia, Colruyt, Securex, Vandemoortele and others, Tekst is now running inside the systems they already use — SAP, Salesforce, Microsoft — without a re-platforming project, without a year-long implementation, without the legion of consultants the old playbook demanded.

This is the difference Process Intelligence makes. The agents work because the layer underneath them finally understands the business.

Quote-to-cash: our beachhead

We've chosen quote-to-cash as our first focus. Not because it's easy — it is famously not — but because it is exactly the kind of process where Process Intelligence pays off.

It is critical: it is, quite literally, how the enterprise gets paid.

It is complex: every quote, every order, every contract is a small negotiation between systems, people and customer-specific rules.

And it is still, in the vast majority of large enterprises, run largely by hand. Tens to hundreds of people retyping data from PDFs, chasing approvals over email, checking contract clauses by hand, escalating exceptions on instinct. No one grows up wanting to retype order data for a living. We built Tekst so they don't have to.

If we can crack quote-to-cash with Process Intelligence, the rest of the back office follows. Procure-to-pay is the next critical process on our roadmap, with claims handling and customer operations sitting right behind it.

Why now

Three things had to be true for Process Intelligence to become possible. All three are now true.

Foundation models can finally read and reason about unstructured business communication — emails, attachments, exception notes — at near-human accuracy. That was not true two years ago.

Enterprises have lived through the first wave of agent demos and learned the hard lesson: agents without process context don't survive contact with reality. The market is now actively looking for the missing layer.

And the cost of not automating the back office has gone from "manageable" to "competitive disadvantage". The companies that figure this out first will run leaner, respond faster, and serve customers better than the ones still running on inboxes and spreadsheets.

The window is open. We intend to walk through it.

What the $13.5M is for

Three things. We're doubling our team from 35 to 70 people by the end of 2026, with the majority of those hires in engineering and go-to-market. We're accelerating product — deepening the Process Intelligence platform, expanding the agent capabilities that sit on top of it, and extending our integrations across the systems enterprises actually run on. And we're going international, building out the customer-facing teams that let us serve global enterprises the way global enterprises expect to be served.

A note to the people building with us

To our customers: thank you for trusting us with the processes that actually matter. The roadmap ahead is shaped by what you've taught us.

To our team: this round is a vote of confidence in what you've built and what you're going to build next. The category we're creating exists because of your work.

To our investors — Elephant, Entourage, and everyone who has backed us along the way: thank you for seeing the shape of this category before most of the market did.

And to the operators, builders and leaders who feel the same frustration we do — that AI should already be doing more, that the back office should already be running better, that the gap between the promise and the reality of agentic AI should already be closed — we'd love to talk.

The agentic era doesn't begin with smarter models. It begins with software that finally understands the business.

That software has a name now. It's called Process Intelligence. And we're just getting started.

‍