The AI readiness framework: How to measure AI ROI in accounting

Why traditional ROI metrics fall short for AI

Here's where most firms get stuck — they try to measure AI using the same metrics they've always used. Billable hours. Time saved on individual tasks. Cost per return.

Those metrics aren't wrong, but they're incomplete. They miss the bigger transformation happening.

Chris uses an analogy that hits home. "Think of it like email for a second. Previously, it took three to five days to get something from desk one to desk two. Now it takes no time whatsoever. Did that make mail irrelevant? Did that make U.S. mail cheaper? No. It changed our behaviors with respect to the way we engage."

AI is doing the same thing. It's making old tasks cheaper and enabling entirely new capabilities.

"If you're worried about the debits and credits and the software doing the debits and credits and your job's going to be irrelevant, you're looking at the wrong thing," Chris explains. "Somebody still has to translate that into human speak. Since the beginning of GAAP, a business owner has said, I don't know what the profit and loss means. You know how I know? Because I have a whole client base full of them."

The real transformation? Moving from reactive compliance to proactive advisory services. From historical reporting to forward-looking guidance. From "here's what happened last year" to "here's what you should do next quarter."

That's the ROI traditional metrics can't capture.

Assessing your current state for accounting AI adoption

You can't build a strategy without knowing where you're starting from. Every firm sits somewhere on the AI maturity curve, whether they realize it or not. The key is an honest assessment of your technology infrastructure, your culture, your leadership commitment, and the barriers that might be holding you back. This isn't about comparing yourself to others; it's about understanding your unique position so you can chart the right path forward.

Where is your firm on the AI maturity curve?

Before you can measure ROI, you need to know where you're starting from. Most firms fall into one of five levels:

Here's something that might surprise you — even if your firm says "we don't use AI," your team probably does. Faba Daniel, Senior Manager in Thomson Reuters product organization and former public accounting professional, calls this out directly. "If a firm chooses not to create a policy or not to engage with AI, it doesn't mean your folks are not utilizing it. This concept of 'we're just not going to use it and we're going to stay the way we've always been' creates this unregulated, unwatched way of engaging with AI that puts your firm and your staff more at risk."

This issue is the shadow IT problem all over again. The solution? Get ahead of it with clear policies and purposeful adoption.

Infrastructure and cultural readiness assessment

Faba, who spent a decade in public accounting before joining Thomson Reuters, emphasizes that readiness isn't just about technology.

Ask yourself these questions:

• Technology audit. What systems do you currently have? What integration capabilities exist? Tools like Thomson Reuters Ready to Review are purpose-built to work with existing tax preparation workflows, using agentic AI to extract and categorize data so you start at "ready to review" instead of starting from scratch.
• Data readiness. Is your firm data accessible and organized? Can you easily pull client information when needed?
• Leadership commitment. Are partners willing to personally adopt AI? This one is non-negotiable. As Faba notes, "Partners must personally use AI tools, this isn't delegate-able. This is firmly like getting partners into the tools. This is a moment, a shift in time that truly transforms the industry."
• Staff openness. Research shows that professionals are increasingly seeing AI as having a major impact on jobs, billing and revenue, and even the need for legal or tax and accounting professionals. That fear is real and needs to be addressed head-on.
• Budget allocation. AI tools aren't free. But as Faba points out, "These aren't cheap tools. But this is going to be the difference between us sustaining and creating a workflow that makes sense for us or constantly chasing this kind of balance that we've been chasing for years."

Identifying your firm's unique barriers

Research from Thomson Reuters examining how AI is reshaping tax, accounting, and other professional sectors consistently shows governance concerns ranking as the top priority for firms. Somewhere between 77% and 80% cite this as their primary hesitation.

Other common barriers include:

• Cost considerations and budget constraints
• Accuracy and reliability concerns with complex tax scenarios
• Client acceptance and consent frameworks
• Lack of clear standards or procedures

Ryan suggests starting with what he calls the "find what you hate" methodology. "Start with pain points, not possibilities. Ask your staff: 'If I never did this again, I would be content.' Those are the tasks you want to tackle first."

The key is scalability. Will automating this process benefit one person or the entire firm? That's the question that separates low-impact experiments from transformational changes.

Building your tax firm AI strategy

A visible AI strategy isn't just a nice-to-have. It's the difference between random experimentation and measurable ROI. But building that strategy requires more than enthusiasm. It demands clear governance, thoughtful decisions about which tools belong where, and a framework that your entire team can rally around. The firms seeing 3.1x ROI advantages aren't winging it. They're following a deliberate approach that balances innovation with responsibility.

Establishing AI governance policies

Let's get one thing straight — governance isn't about restriction. It's about enabling confident adoption.

Chris, who currently employs 13 AI agents across his firms, starts every conversation about AI with professional ethics. "You've always got to start with what's permitted. As tax and accounting firms generally, there's Circular 230 or some professional ethics layer that goes with your license that creates a presumption that you will safeguard data."

The framework is simpler than most firms think:

The PII protection rule. Does the data contain personally identifiable information (PII)? If yes, use a closed system. If not, you have more flexibility, but you should still take reasonable steps to protect client data.

"If you cleanse your data to strip out the PII, the personal identifiable information that's out there, the numbers on the return are not private numbers. It's the Social Security number, the names, all that stuff," Chris explains. "You could screenshot the body of a return, feed it to whatever you wanted, and it would generate results. But if you take extra steps to cleanse the database, like using a client ID rather than all the other identifiers, you're protected."

Client consent matters. Chris adds another layer. "If you want to be a heavy AI adopter and you want to use all the systems available, just get client consent. Have them sign off in the engagement letter that you're going to use it. Clients can permit you to do that as long as you have the appropriate disclosures in place. A lot of our young gun clients want us in this space. They're pushing us in that direction."

Understanding closed vs. open AI systems

Not all AI is created equal, and understanding the differences can save your firm from costly mistakes — or worse — ethical violations.

Closed systems. Purpose-built tools like Thomson Reuters CoCounsel Tax operate in a protected environment. They're integrated with authoritative content like Thomson Reuters Checkpoint, trained specifically for tax work, and designed with PII protection in mind. "Best for research, return review, and compliance tasks with PII," Ryan notes.

General-purpose systems. These options include ChatGPT, Claude, and Gemini. "If you're going to use a general-purpose system trying to get an answer to a tax question, you're going to get bad data," Ryan cautions. "They're trained on the broader part of the internet. If you've done any research online, you know there are things out there that are just wrong."

But that doesn't mean general purpose tools don't have their place. Ryan uses them regularly. "For creating Excel formulas or writing a Python script to process files in a certain way, you're not going to get that out of a closed system. But you can use those general-purpose systems to directly modify Excel spreadsheets in complicated ways."

Enterprise business AI. Microsoft® Copilot and similar tools sit in the middle ground. They offer enhanced security but require verification. Always check for SOC 2 compliance and whether they meet IRS 7216 requirements.

Chris adds an important reminder. "Do not confuse the clickbait from the outside. All of these systems have different branding. You've got to go to the next layer and study: What does enterprise mean? Is your data actually private? That's your responsibility to figure out."

The 3.1 times ROI advantage: What visible strategy delivers

Research shows that organizations with a formal AI strategy are more than three times more likely to realize positive ROI than those without one. But what does a visible strategy actually look like?

It comes down to four levers:

1. Strategy. A documented plan and roadmap that everyone can see
2. Leadership. Executive champions who use AI themselves, not just talk about it
3. Operations. Budget allocation, governance roles, and clear ownership; CIO, COO, and CFO involvement
4. Individual users. Training programs and adoption support at every level

Faba emphasizes the importance of visibility. "That adoption amount that feels like they have a strategic firm-wide approach is somewhere between 16 and 18%. So, 90% of us feel like we absolutely need it. 16 to 18% have really scaled it. The rest are sitting in limbo."

Why does visibility matter so much? "Clear strategy empowers middle management to act confidently," she explains. When staff know what's approved, what's encouraged, and what guardrails exist, they stop hesitating and start innovating.

How to measure AI ROI in accounting beyond billable hours

Here's where most firms get stuck; they know AI is supposed to deliver value, but they're measuring it with yesterday's metrics. Billable hours and time-per-task calculations miss the bigger transformation happening. AI is enabling entirely new capabilities, opening doors to engagements you previously had to decline, and fundamentally changing what your firm can offer. To capture that value, you need a new measurement framework.

Redefining metrics: Time savings and capacity expansion

Let's get practical. Here's the basic calculation every firm should start with:

Hours saved formula:

• Baseline. Hours spent on task before AI
• AI-enabled. Hours spent with AI
• Savings. (Baseline - AI-enabled) × Staff rate × Frequency per year

Example. A research project that took five hours now takes 1 hour with CoCounsel Tax. That's 4 hours saved. At $150/hour, done 20 times per year, that's $12,000 in annual savings from one use case alone.

But here's where it gets interesting. Capacity expansion isn't just about doing the same work faster.

Ryan breaks it down. "The number of clients that we can take on and service shoots through the roof. Things I used to have to turn away, especially complicated engagements where I'd have to do special allocations, track book capital accounts, and large calculations, I can now get done with AI and a spreadsheet in a very short period of time."

Those aren't just efficiency gains. They're new revenue opportunities that didn't exist before.

Metric to track. Revenue from engagements that would have been declined pre-AI.

Service diversification and engagement expansion

Thomson Reuters Ready to Advise solutions are built specifically for this transformation to help firms expand beyond compliance into high-value advisory services.

The shift Chris describes is powerful. "We are going to have real-time data available faster than we've ever had before. We can do a tax estimate based on everything that tax and accounting people have dreamed their clients would give them. Isn't that exactly what we asked for?"

When you can analyze complete data sets in real time, you move from historical reporting to forward guidance. From "here's what happened" to "here's what you should do next."

Metrics to track:

• Percentage of clients receiving advisory services, not just compliance
• Average engagement value increase
• Number of services per client; engagement expansion rate
• Advisory revenue as percentage of total revenue

Ryan emphasizes complexity capability too. "With the enhanced tax research I can do through CoCounsel, I can take on a ton of engagements I would have turned away before. Special allocations, book capital accounts; things that would take hours before."

Those complex engagements? They command premium pricing. That's a measurable ROI.

Identifying high-impact use cases

Not all AI use cases are created equal. Some options will only benefit one person in one engagement. Others will transform your entire firm's workflow. The trick is knowing which battles to fight first, starting with tasks that are annoying enough to motivate change, scalable enough to matter firm-wide, and manageable enough to implement successfully. The right methodology makes all the difference between pilot programs that fizzle out and transformations that stick.

The "find what you hate" methodology

Ryan's approach is refreshingly practical. "Find something that you hate. Ask your staff: 'If I never did this again in my public accounting career, I would be content.' Pick that and see if we can break down those components."

But don't just pick any annoying task. Apply three filters:

1. Scalability. Will this benefit one person or the entire firm?
2. Impact. How much time or frustration does this task create?
3. Risk. Where does this sit on the PII or compliance spectrum?

Faba adds an important insight from her public accounting days. "I was that tech nerd who would spend days building automated spreadsheets. Once I did it, I had the coolest thing and could use it for all my clients. But I was the only one who could do it, the only one who could update my formulas. We're in a different age now. We want to create tools and resources that are scalable."

The difference with AI? You can create those scalable solutions faster, and you can document them so they're transferable.

Mapping the human-in-the-loop framework

Chris offers a simple rule: "How can you use AI so that you spend less time on clicking buttons and more time on human judgment? Those are your two measuring sticks."

The workflow mapping rule: "If you can't map out a workflow with five or six sticky notes on the wall, it's not really a refined workflow. There are too many steps, too many variables. You need to come back to a human-in-the-loop element and maybe break that process down into different components."

Where AI assists:

• Data gathering and organization
• Initial analysis and pattern recognition
• Calculation execution
• Document summarization
• Research compilation

Where judgment matters:

• Client impact decisions
• Strategic recommendations
• Nuance interpretation
• Final review and sign-off

Ryan emphasizes: "The decisions that are going to impact our clients is obviously the number one spot where we've got to be double-checking everything. The AI is just a way to get you to that output faster."

Priority use cases for tax and accounting firms

Tier 1 - Immediate adoption; lowest risk, highest impact:

• Tax research with purpose-built tools like CoCounsel Tax
• Document summarization and review
• Workflow documentation using dictation tools

Ryan calls out the research tool advantage. "When you're asking questions with CoCounsel, you're getting answers way faster, but you're still verifying: Where is that answer getting pulled from? Are there any nuances in the true editorial information?"

Tier 2 - Pilot implementation; moderate risk, high value:

• Return review and accuracy checking with tools like Ready to Review
• Microsoft Excel formula creation for complex calculations
• Client communication templates

Tier 3 - Strategic scaling; higher complexity:

• Automated workflow agents
• Real-time data analysis systems
• Integrated advisory platforms

Chris, with his 13 AI agents, is already living in Tier 3: "The reality in this space is it actually depends on where the firm is and maybe also where your client base might live. Each professional needs to take a step back and recognize where we are in our life cycle of technology."

Building your phased adoption plan

Strategy without execution is just theory. The firms achieving ROI from AI aren't trying to transform everything overnight. They're following a phased approach that builds confidence through small wins, scales what works, and continuously refines based on real-world results. The timeline matters less than the discipline — document first, pilot with intention, and scale with evidence.

Phase 1: Documentation and workflow mapping — months one to three

Start by getting your processes out of people's heads and onto paper or screens. Here's where AI can help immediately.

Ryan suggests. "Using dictation tools is insanely helpful. The rate I can speak at is far faster than the rate I can type. Push a button on your keyboard, speak freely, and have a six-page paper about your processes typed out without having to do it manually."

Phase 1 checklist:

• Document current state; use AI dictation tools
• Map existing workflows visually
• Identify bottlenecks and inefficiencies
• Create AI usage policy
• Define closed versus open system guidelines
• Establish client consent processes
• Implement PII protection protocols
• Select two to three pain point processes for pilots

Faba emphasizes getting those SOPs documented. "It's almost documenting the things that we never wanted to document. In public accounting, we taught staff by them watching us. AI isn't watching us. It's not watching our thought processes."

Chris adds, "Even though AI can help as you build it, write it down, track it, feed the inputs. This is still a game of garbage in, garbage out. If nobody knows how you built it, how do they fix it?"

Phase 2: Pilot implementation with measurable benchmarks — months four to six

Before you start:

• Document current time and cost for pilot processes
• Establish quality metrics
• Set success criteria in advance

During the pilot:

• Start with champion users who are excited about the technology
• One workflow at a time
• Heavy testing and monitoring
• Weekly check-ins to adjust based on real-world results

Ryan's advice is, "You must take the time to try some form of adoption. Pick some low-hanging tasks, try it out, see how many hours it saves your staff. There's risk in every investment you make in business."

Not everything will work. That's OK. As Ryan notes, "'Shelf' ideas that don't work now; revisit in six months. What we couldn't do six months ago might be possible today."

Faba agrees. "The tools and capabilities were vastly different even eight months ago than what they can do today. We'll probably be having a very different conversation of what AI can do a year from now."

Phase 3: Firm-wide scaling and continuous optimization — months seven to twelve

Expansion strategy:

• Train remaining staff on proven workflows
• Share success stories and ROI data internally
• Address resistance with evidence, not theory
• Create AI champions at every level

Continuous improvement:

• Monthly ROI measurement and reporting
• Identify next-wave opportunities
• Stay current with evolving capabilities
• Regular "what we couldn't do six months ago" reviews

Timeline and resource allocation

Leadership time investment:

• Months one to three — five to 10 hours per week
• Months four and beyond — two to five hours per week ongoing

Budget considerations:

• Tool costs; purpose-built tools like CoCounsel Tax, Ready to Review, and Ready to Advise from Thomson Reuters
• Training and change management
• Potential consultant support

Expected breakeven. Most firms see positive ROI within six to 12 months, but remember, you're not just measuring time savings.

The investment imperative. As Chris puts it, "Anything worthwhile is going to be painful and it's part of it. My challenge to the audience is this is really about the journey. How can we make all of this seamless in what it is that we do to make our lives better, our clients' lives better, and our teammates' lives better?"

Managing risk, trust, and governance

All the efficiency gains in the world mean nothing if you violate professional ethics or lose client trust. The good news? Managing AI risk isn't fundamentally different from managing any other technology risk. It just requires updated thinking about where client data lives, how you verify outputs, and what transparency looks like in an AI-enabled practice. The foundation remains the same — protect client data, maintain professional judgment, and stay current with evolving capabilities.

The professional ethics foundation

Chris brings it back to basics. "Circular 230 requirements remain unchanged. Safeguarding client data is paramount."

The cleansed data approach — using client IDs instead of personal identifiers — gives you flexibility while maintaining ethical standards. Insurance and liability considerations remain the same as they always have been.

Building trust through transparency

Client communication about AI usage doesn't have to be complicated. Update engagement letters, include consent language, and explain your human-in-the-loop approach.

Here's something interesting Chris has noticed — "Young clients are often pushing firms toward AI adoption. They want you using these tools."

Continuous monitoring and quality control

Ryan emphasizes the verification requirement. "When we're doing research with AI, I definitely think the human has to be reviewing the ultimate output. AI is just a way to get you to that output faster."

Regular accuracy audits of AI-assisted work, verification protocols for client-impacting decisions, and staying current with AI limitations are table stakes.

Remember to re-evaluate tools regularly. "What we couldn't do six months ago, we can do today," Ryan reminds us. That means both capabilities improve and new risks may emerge.

The investment imperative

The era of early AI adoption has passed. Today marks the strategic phase of AI, in which organizations redefine workflows, reshape value, and build AI directly into the foundation of their business strategy.

The firms that invest now are shaping the future of the profession. The ones that wait? They're facing an acceleration gap that becomes exponentially more costly to close.

The ROI reality is clear — organizations with a formal AI strategy are more than three times more likely to realize positive ROI than those without one.

The practical path forward starts today:

1. Assess your current position honestly. Which of the five maturity levels describes your firm right now?
2. Document one hated process this week. Use dictation tools. Get it out of your head and onto paper.
3. Set aside leadership time for hands-on AI adoption. This isn't delegable. Partners must personally use these tools.
4. Build your visible strategy. Document it. Share it. Make it real for everyone in your firm.
5. Measure ROI across all four categories. Faster, better, talent, and financial. Don't just track time savings.

Chris leaves us with the reminder — "If we don't change our profession for the better, no one else will. Get off your tail, own this stuff, and figure out ways that we can work 40-hour work weeks instead of 60, 80, or 100."

Faba adds the urgency. "There will be firms, competitors, practitioners who are investing the time in doing it. And then we will be sitting in a position where our business is at risk because we can't keep up. They've accelerated the learning too fast and the amount of investment you will need to do then is much harder."

The time to act is now. The framework is here. The tools exist.

The only question left is if you will lead the transformation or scramble to catch up?

Start your AI readiness journey with Thomson Reuters

The tools to measure and maximize your AI ROI are available now. Thomson Reuters offers purpose-built solutions designed specifically for tax and accounting professionals:

• Ready to Review. AI-powered tax prep, human-powered strategy. Leverage agentic AI agents to extract and categorize data, producing tax returns ready for your professional review. Start at "ready to review" instead of starting from scratch.
  
  
• CoCounsel Tax. Your agentic AI platform for tax research, analysis, and workflow automation. Get citation-backed answers and automate complex tasks while maintaining the human expertise and judgment your clients rely on.
  
  
• Ready to Advise. Expand beyond compliance into high-value advisory services. Transform your practice from reactive tax preparation to proactive strategic guidance that drives measurable client outcomes.

Your next step. Contact Thomson Reuters to discuss how these solutions can accelerate your firm's AI readiness and deliver measurable ROI across efficiency, client experience, talent retention, and revenue growth.