Stop Trying to Cram AI Into Old Workflows

A challenge we see nearly every organization struggle with is taking time to reimagine how work gets done when generative AI becomes part of the workflow. Too many organizations want to wedge AI into existing processes, and that causes problems. Most importantly, it limits the impact AI can have.

That's a big reason we see smaller companies and earlier‑stage teams moving faster with AI. They have fewer entrenched processes, fewer legacy systems, and fewer sacred cows, which makes it much easier to redraw workflows around what these tools are actually good at.

The data backs this up. In Wharton's 2025 AI Adoption Report, roughly three‑quarters of companies already report positive ROI from AI investments, and the gains are even higher among smaller organizations that can adapt quickly.

As we take more companies through AI Quest—now more than 50 organizations and thousands of people—we see the same fork in the road over and over again.

Two very different questions

When leaders talk about AI adoption, they're usually circling one of two questions, even if they don't say it out loud.

1. How do we cram this technology into our existing workflows?
2. How do we change the way we think about getting work done to maximally leverage generative AI?

The first path feels safer. You keep your org chart, your processes, and your approvals mostly intact and look for places to "plug in" AI, usually as a faster version of something you're already doing.

The second path is harder. It asks you to question defaults: which steps exist only because we didn't have these tools before? What would this workflow look like if AI were at the center instead of bolted on at the edges?

I understand why most teams reach for the first option. It's less disruptive. But I genuinely believe the second is the only choice for companies that want to come out of this ahead.

What the Wharton study is telling us

One helpful thing about the Wharton report is that it puts numbers to trends we're already seeing inside organizations:

• Usage has crossed into the mainstream. Roughly 46% of enterprise decision‑makers say they use generative AI daily.
• Weekly usage is now routine. More than 80% of enterprise leaders say they use generative AI at least weekly.
• In good orgs, ROI is being measured, not guessed. About 72% report tracking structured, business‑linked ROI metrics like profitability, throughput, and productivity.
• Returns are already material. Nearly three‑quarters of organizations report positive ROI today, and around four in five expect positive returns within the next two to three years.
• Investment is accelerating. Roughly 88% of organizations expect to increase generative AI spend in the next 12 months.

Generative AI isn't going away. It's clearly here to stay. But the question is whether you're going to build the kind of workflows that can capture those returns and become a leader, or spend the next few years trying to retrofit AI into processes that were never designed for it. And very likely fall further behind.

What reimagined workflows actually look like

When we talk about "reimagining work," we're not talking about replacing teams with robots. We want to design workflows where AI does the things it's uniquely great at—things like summarizing, drafting, pattern-spotting, exploring options—so that people can spend more time on judgment, relationships, and complex decisions. The stuff we're uniquely great at.

In practice, that often means:

• Turning multi‑step, copy‑and‑paste heavy processes into single prompts that generate drafts, data, and context for a human to review.
• Moving from "everyone figures it out on their own" to shared prompt libraries and templates that show how your best people already work.
• Designing roles and handoffs around AI‑assisted work. For example, shifting junior team members from doing all the manual production to reviewing, editing, and escalating what AI produces.

Those kinds of changes don't fit neatly into what you're doing today. They require imagining how you may work tomorrow.

Where to start if you're an established company

If you're leading a more established organization, the good news is you don't have to rewrite everything at once. In our AI Quest work, the teams that make the most progress tend to do three things well:

• Design the "AI‑first" version of that flow. Ask, step by step, what changes if you assume an AI assistant is always available: what gets automated, what gets simplified, and where humans add the most value.
• Pick one meaningful workflow, not a thousand tiny ones. Start where AI can make a visible difference to a real team. Often, this is stuff like a proposal process, a content workflow, a customer support task.
• Measure like you mean it. Track throughput, cycle time, error rates, and satisfaction before and after. If you can't show a meaningful change, we need to consider if we're approaching this correctly.

From there, the work is about repetition and learning. We can expand to adjacent workflows, refine the guardrails, and give more people hands‑on experience so they can see what's possible.

The real competitive advantage

Generative AI is going to keep evolving, but the organizations that benefit most won't just be the ones using the latest models. They'll be the ones that were willing to do the hard work of redesigning how work gets done.

If you're ready to move beyond experimenting at the edges, the shift starts with a simple but uncomfortable move: stop trying to cram AI into old workflows, and start asking what those workflows could look like if you built them for this moment from the ground up.