Dan Rogers and StackAI’s Toni Rosinol

AI agents promise to transform enterprise productivity by automating routine workflows. But before an enterprise can reap the benefits of this promised bonanza, they must first identify the workflows that are ripe for automation — and that’s not as easy as it sounds, argues Dan Rogers, CEO of Asana. He tells me:

CEOs say to me, one of the biggest blockers is, I don’t even have a list of my workflows. I don’t have a visual representation of my workflows, and actually, my employees and teams that are working within workflows almost can’t self-realize what that workflow is.

In other words, even though employees know what steps they have to take to complete a regular task — collect some data here, send a report there, get approval from whoever — no one has explicitly mapped it out as a repeatable process. It’s just part of the tribal knowledge of how the organization gets things done. Making these workflows explicit is one of the outcomes of Asana’s acquisition of StackAI, which it completed in May. The tool provides a visualization of complex, multi-step workflows and connects into a wide range of enterprise applications, databases and other resources to orchestrate their completion. He goes on:

I think it’s going to be a game-changer when people start to play with Stack AI, because it actually brings workflows to the rest of us. I think that’s probably why some of the first-generation automation companies didn’t do all that well. It’s because you needed to be an expert — you needed to understand your workflow.

But now imagine you can do workflows in human-speak… With our work graph, you can say, ‘What is the workflow that we use here? What are the steps that we seem to follow every time? Can you help me visualize it?’ And it will just do that lovely visualization.

Accelerating the integration roadmap

Asana already has an established agentic workflow builder called AI Studio, which customers will continue to use to build automations within Asana. StackAI brings strong capabilities in co-ordinating more complex workflows that cut across multiple systems. While these capabilities were on the roadmap for AI Studio — which has already added integrations into productivity apps such as Gmail, Outlook, Slack, HubSpot, Figma and Canva — acquiring StackAI brings forward capabilities that were previously as much as a year out, says Arnab Bose, Asana’s Chief Product Officer. He comments:

They have hundreds of integrations, and they’ve also been able to successfully prove their product-market fit by selling into a lot of customers who are in regulated industries like financial services and healthcare and life sciences. They have these pre-built templates for agentic workflows that are multi-step processes, like a know-your-customer workflow for a financial services firm, or onboarding customers for a healthcare and lab sciences firm, and so on so forth. So what it’s bringing to Asana is this ability to build custom agents in a no-code way and orchestrate complex actions across multiple third-party systems…

When you put Asana plus StackAI together, you’re getting all of the end-user experience and context benefits of Asana, and you’re able to trigger these multi-step orchestrations built in StackAI.

StackAI also brings its own team of Forward Deployed Engineers (FDEs), who play a key role in helping customers figure out where agentic automation will be most impactful. Bose says:

That team is actually quite important in not only doing proper discovery when a customer is buying Stack AI — we want to be certain that the customer has a well-defined need, and that even in the purchasing process we are defining that need — and we stay with them through to adoption…

I honestly think this is required in this day and age because you have to do the discovery work and the consultative post-sale success work with customers to help them re-imagine their business-critical workflows as truly agentic workflows. I’m sure at some point in time, like in two or three years, once this becomes a muscle that everybody has, you’ll need less of it. But today, you definitely need it.

My take

As I mentioned in my earlier write-up of Asana’s recent announcements, AI agents are increasingly connecting across multiple functional domains. The acquisition of StackAI is a response to that trend, significantly advancing Asana’s ability to connect and orchestrate across third-party applications. It’s also symptomatic of a need for a different style of integration than we’ve seen in earlier generations of technology, which means that the traditional API-based approach is no longer enough on its own. Therefore we’re seeing established vendors snapping up startups like StackAI that have specialized in agentic orchestration. I asked Bose about why this demands a different approach to integration and his reply was illuminating:

It’s an outcome-based approach, where you set a goal, and then you can leverage AI capabilities under the covers to determine what is the best possible path to get there. Whether it’s writing code, whether it’s doing something deterministic, whether it’s custom-building a connector.

So I think the flexibility has increased 1,000% and the sort of work that the end user has to do to catch all of the exception cases and think through all the possible if-this-then-that statements no longer exist. They can simply define the outcome or goal, and these new agents are just goal-oriented. They will keep working in a long-horizon way until that goal is achieved. So it’s a very different way of approaching the problem versus deterministically stepping through the entire algorithm and trying to define it.

Asana’s historic inclusion of goals and outcomes in its work graph gives it a head start in this new world, where clearly defining the desired end point is much more important than documenting every single step required to get there — indeed, over-specifying the individual steps may actually handicap the AI’s ability to find the best path. And as Rogers points out, many organizations have taken a similar approach to letting their employees organically figure out the best path to getting the job done, with the result that many of these crucial workflows have never been formally documented. No wonder enterprises need teams of FDEs to help them discover where AI agents can have the most impact. People often talk about technology debt, but I have a feeling that process debt — the accumulation of unnecessarily obtuse and redundant workflows — is an even more entrenched phenomenon. Let’s see if AI can finally help enterprises cut through all of these encumbrances.