The CFO as AI Pioneer: What Adobe's Finance Lab Teaches Us About Measurable ROI

You're sitting in a finance planning meeting, and someone asks: "So, when are we adopting AI?" The question hangs there. You know AI is happening everywhere—product, sales, marketing—but finance? Finance is already busy. You're processing invoices, reviewing contracts, answering stakeholder questions. The last thing you need is another initiative that sounds promising but might create more work.

This is the paradox most finance leaders face: AI feels important, but the path from "sounds interesting" to "actually saves time" feels unclear, especially when accuracy is non-negotiable. One wrong payment or missed contract term isn't a learning opportunity—it's a compliance issue.

Adobe's CFO Dan Durn decided to solve this differently. Instead of waiting for the perfect AI strategy, he turned finance itself into an AI lab. The results tell a story worth learning from.

The Problem: Three Layers of Friction

Here's what Durn identified. Finance teams are drowning in manual work that AI was specifically designed to handle, but the handoff between finance and technology teams creates delays. You submit a request to IT. IT talks to security. Security has questions about data governance. Meanwhile, your team is still manually reviewing contracts or responding to routine emails.

The external problem is obvious: manual processes are slow. Responding to vendor inquiries, processing contract terms, analyzing procurement documents—these take real time away from strategic work.

But there's a second, internal problem that doesn't always make it into presentations: frustration. Your team feels like they're falling behind. Marketing has generative AI tools. Product teams are building with LLMs. Finance? You're still copying and pasting contract excerpts into spreadsheets. That gap breeds a feeling of being left behind, of not being empowered to compete for innovation within your own company.

And beneath that sits a philosophical problem: "Shouldn't we have to do this manually?"

Durn's insight was simple but powerful: don't wait for technology teams to figure finance out. Build it yourself. Move finance from a user of AI (waiting for tools) to a builder of AI (making tools work at finance speed).

How Adobe Did It: The Three-Layer Approach

When I've worked with finance organizations serious about AI adoption, I've seen success cluster around three technical layers: document processing infrastructure, data platforms that unify contract and financial records, and agentic AI—systems that don't just analyze, but act.

Adobe's lab attacked all three layers at once, which is why the results are remarkable.

Layer 1: Document Processing & Automation. Durn's team started with the highest-friction work. Twenty thousand emails come into 19 different finance inboxes monthly. Routine questions. Routine approvals. Routine responses. They built an AI system to auto-respond to structured vendor inquiries, saving over 5,000 hours annually across those 19 inboxes. Not by ignoring emails, but by handling the categorizable ones accurately, freeing humans for exceptions and complex conversations.

This is the speed gain: emails that previously required human triage, reading, and response now get handled in seconds. Humans only touch the emails that need judgment.

Layer 2: Data Unification & Contract Intelligence. Here's where Adobe's approach gets interesting. Finance and procurement teams sit on goldmines of contract data. Terms, conditions, payment schedules, renewal dates, exclusivity clauses—all locked in PDFs and disparate systems. Adobe deployed agentic AI with what they call PDF Spaces, giving finance teams the ability to query entire contract repositories for specific terms in seconds. Need to find all contracts with 30-day payment terms? Done. Which agreements have auto-renewal clauses? Now searchable in real time.

Forrester validated this: document analysis efficiency jumped 45%. More important, contract review time across finance and procurement was cut roughly in half.

Think about what that means operationally. A task that took a full day now takes four hours. A task that took a week now takes 2.5 days. When you're scaling a company, acceleration like that compounds.

Layer 3: Agentic AI & Leadership Alignment. This is the piece most finance leaders miss. Having good AI tools doesn't automatically mean good outcomes. Durn reorganized finance, IT, and security under single leadership—one person accountable for moving pilots into production. That removes the friction between "security says no," "IT says it's complex," and "finance needs it now." Instead, you get collaborative problem-solving with aligned incentives.

The philosophy guiding this layer is captured in one phrase: "accuracy is non-negotiable." Agentic AI can make errors. The lab tested, validated, and only deployed systems where the error rate was acceptable for the financial stakes involved. It's not about perfection—it's about specification.

The Payoff That Matters: Scale Without Headcount

Here's the temptation: assume Durn fired a third of his finance team with these savings. He didn't. The point wasn't headcount reduction. The point was efficiency that enabled growth without proportional cost increases.

That's a different value calculation. If your finance team is processing +30% more transactions, analyzing +40% more contracts, and doing it without adding 30-40 more staff, you've unlocked leverage. You're not extracting cost—you're extracting scale capacity.

This reframes what CFOs should care about. The traditional view: AI saves labor. The smarter view: AI prevents finance from becoming a rate limiter of growth.

Your Playbook: Three Steps to Replicate This

If you're ready to apply this at your organization, here's the path forward:

1. Start with your highest-friction process—not your most strategic one. Durn began with email routing, not with complex financial modeling. Why? Because high friction creates fast feedback loops. You deploy, you see results in weeks, not quarters. That momentum builds internal credibility and funding for the harder layers. Pick something where accuracy is important but the task is well-defined. Auto-routing, document categorization, contract term extraction—these work well as starting points.

2. Map the three layers and identify which one is your actual bottleneck. Don't assume it's the technology. Often it's governance. Can your security team move at the speed your finance team needs? Is IT accountable for getting pilots live? Reorganizing for speed—even temporary reorganizing around an AI initiative—matters as much as the technology itself.

3. Set the accuracy bar before you build. This is the discipline. Define what "good enough" looks like for each use case. Maybe auto-responses need 99.5% accuracy. Maybe contract term extraction needs 98%. Maybe categorization needs 95%. Once you've set the bar, you can build toward it rather than building something and hoping it meets standards.

The Stakes: Why This Matters Beyond Efficiency

This matters because finance teams face a choice. Adopt AI intentionally, with clear ROI frameworks and accuracy standards—or get disintermediated by it. Marketing leaders are already using AI to forecast demand. Product teams are using it to scale. Sales is using it for forecasting and pipeline management. If finance remains the department that "doesn't use AI," you'll find yourself answering questions you shouldn't have to: "Why doesn't finance have real-time visibility? Why do we wait three days for a contract analysis? Why is our financial planning cycle still three months?"

The success vision is clear: a finance team that uses AI to compress decision cycles, reduce errors, and scale without proportional headcount growth. You go from being "the check-writers" to being "the fast analysts."

The failure case is just as clear: you don't do this, and in 18 months, you're the bottleneck. Your stakeholders have already gone around you, built their own reporting systems, used external contractors for analysis. Finance went from a control function to an irrelevant one.

Durn chose the former. His playbook shows it's possible without a complete technology rebuild—just clear thinking about where friction lives and the discipline to tackle it methodically.

Key Takeaways

• Adobe's CFO didn't wait for an enterprise-wide AI strategy. He built a lab inside finance and proved ROI at scale: 5,000 hours saved, 45% efficiency gains, contract review time cut in half.
• The three layers (document processing, data unification, agentic governance) work together. Tackle them sequentially, not in parallel.
• The payoff isn't headcount reduction—it's scale capacity without proportional cost increases. That's how you keep finance from becoming a growth limiter.
• Start with high-friction, well-defined processes. Build momentum. Then tackle strategic work.
• Reorganize for speed. If your IT and security teams can't move at your finance needs pace, you'll stall. Put one leader accountable for production outcomes.

The CFO as AI pioneer isn't a future concept. It's already happening at Adobe, and it can happen at your organization. The question isn't whether you have the technology. You do. The question is whether you have the clarity to deploy it with discipline.

---

Have you started building AI into finance? What's your highest-friction process? Let me know. I read every reply.

Get insights like this weekly → The ROI Signal newsletter at terrainroiintelligence.com