Skip to content
Blog

What Discovery Calls Keep Teaching Us About Back-Office Work

After dozens of discovery calls with established businesses, one pattern stands out: their work and information live in too many places for AI to help yet.

After dozens of discovery calls with established businesses in many industries, the biggest back-office problem we hear is that the work lives in too many places. Before AI can help, teams usually need to map how work actually moves and get their people comfortable with change.

What do operations leaders tell us first?

Almost never "we need AI." Usually it's that their information is everywhere.

At Soren, an AI-native engineering firm, we've had dozens of discovery calls with established businesses in many fields, from law firms and lenders to healthcare teams and nonprofits. Here's what comes up most, roughly in order:

The back-office problems that come up most in Soren's discovery calls, 2026
What we hearWhat it usually means
"Our information is everywhere."Work and documents live across too many systems
"We don't know where to start."No map of how work flows today
"We're worried about falling behind."Leaders feel pressure to act on AI and are unsure how
"Where does our data go?"Security and privacy have to be answered first
"We tried a vendor before."A tool or partner didn't stick
"Intake takes forever."Work gets stuck at the front door
"A few of us use ChatGPT."Individuals have tried AI; the organization hasn't adopted it

Why does work end up in so many places?

Because businesses add tools one problem at a time, and connecting them is rarely anyone's main job.

Atlassian's 2025 State of Teams research found that teams lose 25% of their time just searching for answers.1 That matches what we hear. The most common surprise in our assessments is how scattered a team's information is. Most teams need their documents and data centralized before automation can do much. That's why one of the most common first workflows we build is a searchable internal knowledge base.

Nobody's problem is AI. It's that the work lives in six places.

Why does security come up so early?

Because these businesses handle other people's money, health and legal matters. They should ask.

It is a fair question with a practical answer. We build inside the systems a business already has, so its data stays where its policies require. Once that is settled, the conversation moves on to what the team wants to get done.

What do teams ask for versus what they actually need?

They ask to get from point A to point B. They need the map in between.

Most teams have real ambitions for AI. They just don't know how to get there, and it's tempting to jump straight to the end. We move fast. With engineers embedded in your team, the first workflow is usually live within two to three weeks of getting access. But getting workflows solid takes a month or two of iteration with the people using them.

It's not instant, and I think it's better to be honest about that up front.

What do these calls tell us about how businesses want to buy AI?

They want the result, and they want getting it to be simple.

Almost nobody on these calls asks for a tool. They describe an outcome: reports that take minutes, intake that doesn't stall, hours back for the team. They want one group to be responsible for it, and they don't want to become AI experts first.

That is why we built Soren as an AI-native engineering firm. A small team of engineers, working with AI agents, takes the goal, builds it into the systems the business already has and stays to run it. I wrote more about this in the diffusion of AI across the real economy.

Does it look different from one industry to the next?

On the surface, yes. Businesses in different fields work differently, measure value differently and care about different workflows.

A law firm might care most about intake, conflicts and billing. A bank might care about credit memos and reporting. A healthcare team might start with prior authorization or patient intake. Even ROI looks different: billable hours recovered, turnaround time on a loan, staff time freed up for patients.

The pattern underneath is the same, though: scattered information, careful buyers, and a team that needs to trust the change. That's why we start from how a business actually works rather than from its industry. You can see the same pattern on our pages for law firms and any other team.

What surprised us about AI adoption?

How low it still is.

Building in San Francisco, it's easy to assume everyone uses AI every day. We do. Our whole team runs on it. But the Census Bureau found that 19.8% of US businesses were using AI as of May 2026.2 Gallup found that only 25% of U.S. employees say their employer has communicated a clear plan for AI.3 And Pew found in 2025 that 52% of U.S. workers are worried about AI's future impact in the workplace, while only 36% feel hopeful.4 It's a big part of why AI pilots stall after the demo.

So if you feel behind, you're in the majority, and catching up takes less than most people expect. We laid out the fastest, lowest-cost way to catch up on AI.

A lot of people are used to how they've done things for years, and they're not eager to change. That's not stubbornness. It's a reasonable response to tools that haven't earned their trust yet.

What should an ops leader fix first this quarter?

Two things, and neither is buying software.

  1. Work on AI sentiment. Help people understand where AI is useful and where it isn't. Make it clear it's there to help them, not replace their judgment.
  2. Map how the work actually runs. Write down where work comes in, where it goes, which systems it touches and where it gets stuck. Once that's mapped, it becomes obvious where AI fits.

McKinsey's 2026 State of AI report found that nearly three-quarters of AI high performers have fundamentally redesigned their workflows, compared with about a quarter of other companies.5

You have to start from the beginning. Build the foundation first. The automation comes after.

Map where your work lives

Start with a 30-minute call with our team. We'll help you map where your work lives and where AI can actually help.

Frequently asked questions

What is the biggest back-office problem at traditional businesses?

In our discovery calls, the most common problem is that work and information live in too many places. Documents, emails and data are spread across systems that don't talk to each other, so people spend time searching and retyping. Centralizing that information usually comes before any automation.

Should we map our workflows before adopting AI?

Yes. Mapping where work comes in, which systems it touches and where it stalls shows you where AI can actually help. Without that map, teams often automate the wrong step or buy a tool that doesn't fit. Companies seeing the most value from AI are far more likely to have redesigned their workflows.

How do we get employees comfortable with AI?

Start by explaining where AI helps and where people still make the decisions. Then involve the team early: interview them about their work, test tools with them, and train them as you go. Comfort comes from seeing AI make their own work easier, not from an announcement.

Sources

Founder & CEO, Soren AI

Kevin Xie is the founder and CEO of Soren AI (YC F25). He studied AI and mathematics at MIT and conducted research at MIT, Harvard and Johns Hopkins, with work published in Nature Biomedical Engineering. Soren is an AI-native engineering firm: its engineers work with AI agents to build AI into the day-to-day work of established businesses, from law firms and financial firms to healthcare organizations and nonprofits.

Let’s map your first workflow.

Get started today