The Four Phases of AI Transformation in Accounting
A Practical Roadmap From Compliance Automation to Advisory Practice
By Daniel Lawrence, Founder, Bots For That
The accounting profession has a problem that is as specific as it is common.
Firms know they need to transform. They have read the white papers, attended the webinars, appointed the champions, and formed the working groups. They accept, at least intellectually, that AI is changing the landscape and that standing still is no longer a viable strategy.
What most of them do not have is a clear, sequenced, practical answer to the questionthat actually matters: where do we start, what do we do next, and how do we know when we’ve got there?
This article is an attempt to answer that question, not in abstract terms, but in a specific, ordered framework that reflects how organisations in other industries have successfully navigated exactly this kind of transformation, and that I have refined through deploying AI in some of the most demanding operational environments I have encountered.
The framework has four phases. They are sequential by design. Each one builds the foundation for the next. And the destination, Phase 4, is not a technology outcome. It is a business model outcome: a practice that is structurally positioned to deliver, and charge for, the advisory work that clients genuinely need.
Before the Phases: The Precondition Nobody Talks About
Before any firm can meaningfully engage with this framework, there is a precondition that must be met. It is not glamorous. It is not the part that gets talked about at conferences. But without it, every subsequent investment in AI delivers a fraction of its potential.
That precondition is data quality.
EY’s global assurance innovation leader put it plainly in their 2026 outlook: the limiting factor for progress in accounting AI will no longer be AI capability, it will be data readiness. The quality, lineage, and availability of financial and operational data needed to power these tools responsibly and at scale.
AI does not compensate for poor data. It amplifies it. A system that reasons across inconsistent, incomplete, or poorly structured data does not produce uncertain outputs, it produces confidently wrong ones. And in a profession where outputs have professional and regulatory consequences, confidently wrong is considerably more dangerous than uncertain.
Before your firm invests in AI tooling, invest in answering three questions. Are your workflows understood clearly enough that a new team member could follow them without asking for help? Is your client data consistently structured across your portfolio, with same fields, same formats, same naming conventions? And do you have clean, reliable connections between your data sources, or are people copying and pasting between systems?
If the answer to any of these is no, start there. Not because it is exciting, but because it is the foundation on which everything else is built.
Phase 1: Compliance Automation
Remove the work that doesn’t need you
The first phase is the unglamorous one. It is also the most important.
Compliance work, with the systematic, high-volume, rules-governed production of returns, reconciliations, reports, and filings, is the engine room of most accounting practices. It is also, by any honest assessment, the category of work that is most directly threatened by AI and the most immediately automatable with tools that exist today.
The goal of Phase 1 is not to reduce headcount. It is to remove the category of work that does not require professional judgment from the professional’s workload, and redirect that capacity toward work that does.
The workflows that belong in Phase 1 share three characteristics.
1. They are high-volume: repeated many times, across many clients, in broadly similar ways.
2. They are well-defined: there are clear inputs, clear rules, and clear outputs. And they are data-dependent.
3. They rely on structured information that exists in digital form and can be reliably retrieved.
In accounting, that description fits a large proportion of the work: bank reconciliation, transaction categorisation, VAT return preparation, payroll reconciliation, document chasing, deadline management, management account production, and the numerous administrative steps that connect them. None of this work requires an accountant’s judgment at every step. Much of it does not require an accountant’s judgment at any step, it requires an accountant’s review or sign-off, which is a different thing.
What Phase 1 (should) look like in practice: A new client document arrives. The system extracts the relevant data, classifies it according to established rules, flags anything outside expected parameters, and routes it to the appropriate stage of the workflow, without a human initiating any of those steps. The accountant’s inbox contains a summary of what arrived, what was processed, and what needs attention, not a pile of unprocessed items to work through.
At the portfolio level, the compliance position of every client is visible in real time. Deadlines approaching, documents outstanding, anomalies flagged, liabilities estimated. Not constructed by an accountant spending time on each client file. Generated continuously, from live data, in the background.
The Wolters Kluwer Future Ready Accountant report found that firms with highly integrated technology stacks, the prerequisite for this kind of Phase 1 automation, are 53% more likely to be experiencing revenue growth than those operating in more fragmented environments. The integration is not the source of the growth directly. But it is what makes Phase 2 possible, and Phase 2 is where the growth comes from.
The honest challenge of Phase 1: Most firms underestimate the effort required to implement Phase 1 properly. Not because the tools are difficult, the tools are largely ready. But because the workflows rely on being more interconnected, which generally relies on being well understood, documented and standardised before they can be automated. You cannot easily automate a process that is different every time someone does it, or that depends on institutional knowledge that lives only in someone’s head.
The firms that rush Phase 1, that implement tools on top of undocumented, inconsistent processes, produce fast, consistent, wrong outputs. The firms that take the time to standardise first, automate second, produce something genuinely transformative.
The effort is front-loaded and significant. The return compounds for years and I’ve witnessed this first-hand with organisations over the past decade in deploying robotic process automation.
Phase 2: Portfolio Intelligence
See everything, across everyone, all the time
Phase 2 is where the nature of the accountant’s role begins to change.
With Phase 1 in place, with the routine compliance work handled systematically and the data flowing consistently, something new becomes possible. The entire client portfolio becomes visible, in real time, not as a series of individual files but as an integrated picture that no accountant could previously construct without spending hours pulling it together.
Portfolio intelligence means knowing, at any moment, which clients have anomalies in their accounts that warrant a conversation. Which ones are approaching a cash position that should concern them. Which ones have compliance deadlines this week and where their documentation stands. Which ones have growth indicators in their data that suggest a strategic conversation is overdue.
This is not possible without Phase 1. You cannot generate portfolio-wide intelligence from data that is not consistent, structured, and up to date. The investment in Phase 1 is also, therefore, an investment in Phase 2 capability.
What Phase 2 (should) look like in practice: On a Monday morning, a partner can see not just their own client list but the compliance position of the entire firm’s portfolio, flagged, sorted, and prioritised by urgency and commercial significance. The system has already identified three clients whose accounts contain unusual patterns. Two clients have deadlines this week and outstanding documents. One client’s data shows revenue growth of 40% in the past two quarters, a signal that they may need a conversation about structure, planning, and what comes next.
None of these observations required an accountant to spend time on a client file. The system generated them. The accountant’s job is to decide which conversations to have, and then to have them.
This is the moment at which the accountant transitions from information processor to information interpreter. And it is the moment at which the firm’s capacity, freed by Phase 1, begins to provide value and even generate new revenues rather than simply being absorbed by the next compliance task.
The honest challenge of Phase 2: The primary challenge here is not technical. It is cultural. Partners who have spent careers working through client files one at a time, building their knowledge of each client through direct engagement with the data, sometimes find portfolio-level visibility disorienting. It removes the familiar rhythm of the work.
The reframe that helps most: the system does not know what the numbers mean. It only knows what they say. The interpretation, the judgment, the context, the relationship, still belongs entirely to the accountant. Phase 2 gives you better information, faster. It does not tell you what to do with it.
Phase 3: Client-Facing AI
More contact, less friction, better relationships
By Phase 3, the firm’s internal operations are substantially transformed. The compliance engine is running efficiently. The portfolio is visible and intelligently monitored. The team
has capacity that did not previously exist.
Phase 3 is about directing some of that capacity, and some of the AI capability, outward, toward the client relationship itself.
This is not about replacing the accountant in client interactions. It is about changing the texture of those interactions, increasing their frequency, improving their quality, and removing the friction from the routine contact that currently consumes time without generating proportionate value.
What Phase 3 (should) look like in practice:
Clients receive automated, personalised reporting on a schedule that suits and benefits them, monthly management summaries, quarterly compliance updates, annual planning packs, generated from live data, formatted to the firm’s standards, and delivered without anyone producing them manually.
When a client has a routine query, a question about a specific transaction, a request for a YTD summary, a question about an upcoming deadline, a conversational AI interface provides an immediate, accurate response from the firm’s own data, in the firm’s own voice. The accountant’s time is not consumed by a question that took thirty seconds to answer.
When the system identifies something worth a conversation, a cash flow concern, an anomaly, an opportunity, a proactive alert goes to the client, signed off by the accountant, from the firm. The client experience is one of a firm that is always watching, always thinking ahead, always adding value. Not one that checks in at deadline time.
The Karbon 2026 State of AI in Accounting report suggested that 98% of firms now use AI, with the majority using it daily or several times a day, but only 21% have an AI policy or strategy. Phase 3 is where the absence of strategy becomes most visible, because client-facing AI without governance is client-facing risk. Every output that goes to a client needs defined standards, review processes, and accountability. Implementing Phase 3 without that structure is the fastest way to damage the client relationships that the whole framework is designed to protect.
The honest challenge of Phase 3: Personalisation is everything here, and most firms underinvest in it. AI that generates generic client outputs, in generic language, at generic intervals, is not a client relationship tool. It is a mail merge with better technology.
The firms that get Phase 3 right are the ones that invest in training their AI, configuring it to reflect the firm’s voice, the client’s specific context, the nature of the relationship, before they deploy it. The output should feel like it came from the accountant who knows that client, not from a system that processes all clients identically.
This is the missed opportunity that the majority of the market is currently missing entirely. The question is not whether you have AI. It is whether your AI sounds like you.
Phase 4: The Advisory Practice
The destination the profession has always been capable of reaching
Phase 4 is not a technology implementation. It is a business model transformation.
With Phases 1 through 3 in place, something exists in the firm that has likely never existed before at this scale: capacity. Real, substantial, redirectable capacity, hours that were previously consumed by data processing, manual workflows, and routine client contact, now available for the work that actually requires an accountant’s professional judgment, experience, and relationship.
The question that Phase 4 asks is: what do you do with it?
The answer, for the organisations that have navigated this journey most successfully in other industries, is to build a genuinely more value-oriented business model. One that is not structured around the hour, because the hour is not the unit of value that clients actually care about, but around the outcome. What did the advice change? What decision did it enable? What risk did it prevent, or what opportunity did it capture?
A perfect example of this in another unrelated industry that’s doing this successfully, is in the betting and gaming sector, where, having worked through phases 1 to 3, phase 4 could deliver real-time, AI-driven insight to proactively assist shop staff-punter conversations that massively reduced manual effort, errors and delays, and exponentially increased outcomes for all involved.
What Phase 4 (should) look like in practice: The firm offers services that did not previously exist, not because the expertise was absent, but because the capacity was not there. Strategic financial planning. Business model review. Acquisition support. Growth advisory. Tax strategy for the next three years, not just compliance for the last one.
Pricing is restructured around the value of those services rather than the time taken to deliver them. Clients who previously received a compliance relationship receive a strategic one. The nature of the conversations, and the commercial relationship that supports them, is fundamentally different.
The profession has been talking about this transition for years. Cloud got us part-way there. AI does not make it inevitable. But it removes the primary structural barrier, the lack of capacity, that has prevented most firms from making it at scale.
EY’s global assurance innovation leader put it directly: as AI becomes the super-assistant behind the scenes, more accountants will have the capacity to deliver the kind of advisory work clients always hoped for. That is not an aspiration. For the firms that have completed Phases 1 through 3, it is an operational reality.
What Phase 4 looks like outside accounting, and why it matters: The clearest illustration of what Phase 4 actually delivers in practice comes not from accounting, but from an industry that has already completed this journey: betting and gaming.
A major operator I worked with had, through Phases 1 to 3, automated its core transactional and compliance workflows, built real-time portfolio visibility across its estate, and deployed customer-facing tools that reduced friction in routine interactions. By the time Phase 4 arrived, the infrastructure was in place to do something qualitatively different.
The result was a real-time, AI-driven insight layer deployed directly into the conversation between shop staff and customers. The system processed live data, transaction patterns, customer history, market movements, contextual signals, and surfaced relevant, timely information to the member of staff at exactly the moment it was needed. Not after the conversation. Not as a report to review later. In the moment, in the flow of the interaction.
The outcomes were significant across every measure that mattered. Manual effort dropped sharply. Errors and delays were reduced. And the outcomes for staff and customers alike, the quality, relevance, and speed of the interaction, improved exponentially. Not because the staff became more knowledgeable overnight. But because the AI gave them the right information, at the right time, in a form they could act on immediately.
The parallel for accounting is direct. Replace the shop floor with a client meeting. Replace the punter with a business owner who has a question about their options. Replace the market data with the client’s financial history, their current position, their upcoming obligations, and the relevant planning opportunities the data reveals.
Phase 4, done properly, is not the accountant working harder or knowing more. It is the accountant, supported by a system that has already processed everything relevant, having a better conversation, one that is faster, more accurate, more proactive, and more genuinely valuable to the client than anything the compliance relationship ever made possible.
That is the destination. And the path to it, through Phases 1, 2, and 3, is the same in every industry where it has been successfully built.
The honest challenge of Phase 4: The hardest thing about Phase 4 is not the services or the pricing. It is the identity shift.
Accountants who have spent careers defined by technical compliance expertise, who have built their professional identity around being the person who manages the complexity of the tax system on behalf of their clients, sometimes find it difficult to transition into a role that is less about technical execution and more about strategic conversation.
This is a real and legitimate challenge. It deserves to be named, not glossed over.
The answer is not to dismiss it. It is to invest in the transition, in training, in new service design, in the client conversations that test and refine the advisory model before it is fully rolled out. The firms that do this work thoughtfully and systematically will find that their clients’ appetite for strategic advice is considerably greater than the compliance relationship ever suggested.
The Framework Summary
Where to Start
The question I am asked most often, having laid out a framework like this, is: which phase are we in, and how do we move to the next one?
If you want a single, immediate, actionable starting point: document one workflow. Pick the highest-volume, most clearly defined process in your firm, the one that consumes the most time, involves the most repetition, and requires the least professional judgment
at each step. Write down exactly how it works, from trigger to output, with every step and
every decision point.
Then ask two simple questions about each step.
First: could someone follow this step using only the written instructions, without needing to ask anyone for help or apply their own experience to figure it out? If the answer is yes, that step is almost certainly automatable. If the answer is no, if it truly relies on someone’s judgment, their knowledge of a particular client that can’t be written down, or their professional experience to get right, that step still needs a human.
Second: is the information this step relies on already in a system somewhere, in a consistent format, or does someone have to go and find it offline, digitise it, reformat it, or re-enter it from somewhere else first? If the information is already clean and consistent, it can feed an automated process. If it lives in emails, in someone’s head, or in a spreadsheet that is formatted differently every time, that is not a data problem that automation will solve, it is a data problem that needs to be fixed first.
Neither question requires technical expertise to answer. They just require honesty about how the work actually happens versus how it is supposed to happen. That gap, in most firms, is where the real opportunity lives.
That exercise will tell you whether you are ready to begin Phase 1, and if not, exactly what needs to happen before you are.
The four phases are not a destination. They are a direction. The firms that are moving along them, deliberately, sequentially, with clear goals and honest assessment at each stage, are the ones that will lead this profession for the next decade.
The window is still open. But the firms already in Phase 3 are not waiting for you.
Daniel Lawrence is the Founder of Bots For That and the creator of their unique automation operating system and suite of AI-powered tools for the accounting and bookkeeping sector. With over a decade of experience deploying enterprise automation and AI in highly regulated industries, he writes about AI transformation in accounting from the outside in.
© 2026 Bots For That. Part of the Making Accounting AI thought leadership series.
