Adapting with AI Part 2 of 3: Running an Efficient Team in a Market That Won’t Sit Still

Running an Efficient Team in a Market That Won’t Sit Still

“Every new technology we’ve invented–from language, to books, to the mobile phone–has defined, redefined, deepened, and expanded what it means to be human.” – Reid Hoffman, Super Agency: What Could Possibly Go Right with Our AI Future

The conversation around AI is quickly moving away from just an efficiency move to redefining what it means to provide value as a professional firm to our markets. We are having to redefine what it means to even provide accounting and tax services to the market for a fee.

Let me share an example that I often share with others.

I share this to make a point that you’ll have to redefine what it means to provide value. And clients will reimagine what they want from you and your firm. The internal transformation of what we come to believe as accountants and what our services provide (the transformation our clients don’t see) is the hard part. It’s not a software purchase. It’s not the demo. It’s the months of quiet, consistent work of actually changing what a team believes about its value, how the team operates, and the longer, harder work of changing what a team is capable of. Most firm owners understand this in the abstract. What they underestimate is how long it actually takes to perform this transformation.

The Internal Transformation Clients Don’t See

AI adoption inside of a professional services firm follows a predictable arc, and it is not linear. In the first weeks, there’s genuine enthusiasm with the new tools, new possibilities, and a sense that the future is finally arriving. Then comes possible disillusionment. The outputs aren’t quite right. Prompts have to be rewritten. Some team members embrace it; others quietly resist. The wins are real but inconsistent. Around month six, if the firm hasn’t quit, something shifts into selective adoption. Certain tasks, certain people, certain workflows where AI has proven itself start to become the norm. By month twelve or beyond, the new normal starts to take shape.

Some firms quit in the trough. The ones who don’t are the ones who will define what this profession looks like in five years. I’ve consulted on change management efforts with firms for many years, so we knew this was going to be a long haul for our own firm. We started the process by releasing a document to our team at our team retreat called ‘Principles & Practices of a “Human First, AI Forward Firm.”‘ This was the beginning of the change management process.

Part of change management is simply breaking old habits built into professionals over many years.

How do you change? First, start with your beliefs from the leaders and how this will look. Then we put an “AI Learning Lab” GPT into ChatGPT (whom we collectively call ‘Chad’ in our firm) so that our team could simply practice letting go of older ways of thinking, and begin defaulting to the AI tool for their lower-level analysis, email construction, and to remind themselves to stop key punching data when they don’t have to do that anymore. Part of change management is simply breaking old habits built into professionals over many years. This process alone will take a solid 8 months to a year for a full team, since everyone comes along in their beliefs and behaviors at different paces.

What AI Efficiency Actually Means for Your Team

When AI tools reduce manual labor by 20 to 30% (and if done well, over time they can) every firm owner has the same instinct: redirect that freed capacity to higher-value work. More advisory. Deeper client relationships, strategic planning instead of transaction processing, etc.

That instinct is exactly right. But the problem is the gap between knowing where capacity should go and actually getting your team there.

Redeploying freed capacity is a reskilling and change management challenge. It takes most firms 18 to 24 months to fully execute that transition and actually realize the benefits. And that timeline assumes the firm is reasonably good at training and developing their people (most are not, which is why Thriveal has a robust training division).

Weak internal training is one of the most consistent vulnerabilities in small and mid-size accounting firms. It’s not a secret. The work gets done because experienced people figure things out, not because there’s a systematic process for building capability. This has always carried a cost. In the AI era, the cost gets higher because the tools your software partners are deploying can only generate leverage if your team actually knows how to use them. This is why the wins from AI acceleration may arrive more slowly than the technology promises. The bottleneck isn’t the software. It’s the people pipeline. And until firms invest seriously in building that pipeline, in structured training, consistent skill development, and genuine change management, the efficiency gains will remain partial and inconsistent.

The real win that compounds over time is capacity redirection. Time that used to go to processing invoices could now go to client communication, exception review, financial insight summaries, and the kind of strategic thinking that clients actually want to pay for. But that redirection only happened in firms that were intentional about it.

The Prompting Gap

There’s another dimension to team readiness – most teams cannot use AI effectively because they don’t know how to direct it. Related to the generative power of AI, firms need to train teams how to prompt and contextualize the product.

The research is clear on this: insufficient knowledge is the primary barrier to AI adoption, consistently outranking cost and infrastructure as an obstacle. Bad prompts produce outputs that take more time to fix than the original manual work would have. Good prompts produce outputs that are 90% ready with clear signals on the 10% that needs human judgment. The difference between those two outcomes is skill, which can be taught.

These five principles make the difference when prompting and contextualizing generative AI tools:

• Establish role and context. Don’t ask AI to “analyze this.” As an example, ask it to analyze “as a CFO reviewing SaaS company financials, with these specific concerns.” Role definition activates the right domain knowledge. You need to spend time with your generative tool ‘getting it ready’ to work with you on what you need. Generative AI tools ‘know everything’ so you have to narrow the field of what you want to work on with the tool.
• Provide explicit quality criteria. Tell the AI what you want it to prioritize and where to be cautious. “If less than 75% confident, say so explicitly” prevents the confidently-wrong outputs that erode trust (the hallucinations we mentioned earlier). Ask the tool to ask you questions so it can check first if it’s about to head in the right direction.
• Request structured outputs. Structured outputs are easier to validate and integrate. “Provide analysis in a three-column table: Finding | Evidence | Recommendation” takes seconds to specify and saves significant review time. If you don’t like the output, it’s easy to say “The column headings and order don’t make sense. Redo it.” And Claude won’t even get mad at you.
• Use examples to calibrate. Show AI what good looks like. “Here’s a strong analysis from last month–use this structure” trains the output faster than paragraphs of instructions. Give the tool multiple examples to keep it focused.
• Build transparency about uncertainty. Ask AI to flag assumptions and rate confidence. This turns the tool from a black box into a transparent collaborator, and prevents your team from trusting hallucinations.

A second category deserves equal attention – agents embedded directly in your financial software. Unlike standalone tools, these work inside real client data–with a precision and efficiency no external tool can match. Agent Mel from Melio and Jax from Xero, for example, understand natural language and context. You converse with the data rather than search it–surfacing anomalies, spotting opportunities, analyzing payment behavior. Mastering both categories is where the prompting advantage lives.

Leadership in Change Management Is the Real Work

Individual skill definitely matters. But team-level adoption is what determines organizational value. And we’ve already pointed out that our organizational value is changing. Team-level adoption requires leadership, not just access to tools.

The research on change management in technology adoption is unambiguous. Prosci’s benchmarking study of more than 10,800 change practitioners found that projects with excellent change management oversight are seven times more likely to meet their objectives–88% succeed versus just 13% without it. McKinsey’s research quantified what that gap means in dollar terms: organizations with strong change management at every level captured an average of 143% of their expected ROI from major initiatives, while those without effective change management realized just 35 cents on the dollar.

And for AI adoption specifically, McKinsey found that organizations leading in the field share a common trait: they invest twice as much in change management as in building the technology itself. Most firms invert that ratio where they spend heavily on software and minimally on the human work of actually changing behavior. Your success in the adoption of AI into your firm has more to do with your change management efforts than it does with the tool itself. Further, you as the leader, are the accelerant that can really drive this change.

For accounting firms, change management means four things done consistently:

• Regular AI learning sessions. Share what’s working, what’s failing, and what’s been learned. Normalize experimentation and make the results visible. This can be done monthly, bi-weekly, etc.
• A prompt template library. Document successful prompts and make them accessible to everyone. This compresses the individual learning curve from months to weeks.
• A quality review process. Especially in the first 90 days with any new AI workflow, systematically track what AI gets right, what it misses, and what reliably needs human judgment. Build that knowledge into your process.
• Transparent communication about AI use. Tell clients which parts of their work involve AI assistance and what the human review looks like. Clients who understand your process trust it. Clients left to imagine it often imagine the worst.

The Timeline to Expect

Here is what a realistic AI adoption arc looks like for a small firm that commits to doing it well:

• Months 1 to 3: Tool selection and early experimentation. Some wins, some frustration, significant inconsistency. The goal is learning, not efficiency.
• Months 4 to 9: Selective adoption. Specific workflows where AI has proven itself become standard. Team members develop differentiated skill levels. Early prompting libraries emerge.
• Months 10 to 18: Systematic integration. AI-assisted workflows become the default for defined task types. Quality review processes mature. First measurable efficiency gains become reliable.
• Months 18 to 24+: Capacity redirection becomes real. Freed-up time consistently flows toward higher-value work. The firm’s service model begins to reflect what AI actually makes possible.
• Month 24+: With capacity redirection a reality, now you have to consider the need to launch new service lines with new prices–and you want to do this in a way that allows the team to deliver the service, not the owner. This may be where you create new revenue lines without having hired new team members.

That last stage, where the efficiency gains and value redirection actually translate into new revenue or expanded client value, is what my former client Marcus was imagining had already happened. For most firms, it’s 18 months away at minimum (so don’t despair). The firms that start now are the ones who will be having a very different conversation with their clients in two years. The internal transformation your clients don’t see in your change management process is the reason you’re not ready to cut prices. It’s also the reason that your value is about to increase significantly (if you do the work).

Download the full eBook

← Part 1: The Email Every Firm Owner Dreads

Continue reading → Part 3: The Value of Conversation