How Harris, Hardy & Johnstone Cut Advisory Review From an Hour to Five Minutes

Professional services firms usually adopt AI at the edges first. They use it to draft marketing copy, summarize meetings, or clean up email. Harris, Hardy & Johnstone went after a harder target: the billable core. By 2025, the Mississippi-based CPA and advisory firm was using Thomson Reuters’ CoCounsel Tax and Audit to reduce some document review work from about an hour to five minutes, compress technical tax issue checks from 15 to 30 minutes down to roughly one minute, and help win new business within six weeks of adoption. For accounting, legal, and advisory firms, that is the more important AI story: not content generation, but cycle-time compression inside expert work.

The appeal of consulting and compliance businesses has always been straightforward. They sell judgment at a high hourly value. The weakness is just as clear. A surprising share of that judgment still depends on low-yield labor: reading long agreements, pulling out key terms, cross-checking facts against rules, tracing prior guidance, and converting scattered raw material into something a professional can trust. Clients pay for conclusions, but firms spend huge amounts of time building the path to those conclusions. That gap is where margin gets squeezed.

George G. Crowell, a shareholder at Harris, Hardy & Johnstone, described exactly that kind of burden. One example involved reviewing a limited partnership agreement to identify audit-relevant provisions, an exercise that had taken around an hour before AI was introduced. With CoCounsel, the firm reduced that task to about five minutes. Another example came from tax research and issue validation. Questions that used to require 15 to 30 minutes of digging could be narrowed to about a minute for an initial answer path, giving the team a much faster starting point for expert review.

The technique is not “AI writes accounting.” It is better understood as retrieval-augmented professional review. The input is unstructured client material such as agreements, workpapers, tax questions, or prior filings. The processing layer uses document parsing, semantic retrieval, and prompt-driven reasoning to identify relevant clauses, surface authoritative guidance, and organize the material around a professional task. The output is not a final filing or signed opinion. It is a condensed, traceable working draft that lets an accountant or advisor move immediately to verification and judgment. That distinction matters. In professional services, the economic win comes from shrinking the time spent finding and assembling relevant facts, while keeping the expert responsible for the final call.

A concrete audit example shows why this matters. Say a firm receives a partnership agreement before planning fieldwork. A senior or manager has to review the file for allocation rules, special provisions, reporting obligations, and any clause that could affect risk or documentation requirements. Done manually, that means reading page by page, highlighting sections, and building a checklist for the audit team. In an AI-assisted workflow, the reviewer uploads the agreement and asks the system to identify clauses relevant to audit planning, partner allocations, restrictions, and disclosure obligations. The system returns extracted sections, short summaries, and a structured starting map. The human then checks the flagged clauses, corrects anything ambiguous, and uses that condensed output to guide the engagement. The result is not that expertise disappears. It is that expertise starts later, after the document has already been reduced into something usable.

That is why Harris, Hardy & Johnstone’s result is more significant than a simple time-saving anecdote. If a professional can turn one hour of pre-analysis into five minutes, the gain is not only labor efficiency. It changes staffing flexibility, turnaround speed, and the size of work a firm can accept without adding headcount. In service businesses, capacity is usually trapped inside expert attention. AI does not eliminate that bottleneck, but it widens it enough for the same team to move faster across more matters.

The six-week new-business result is also worth noticing. Many firms think of AI as an internal productivity tool, but clients often buy responsiveness as much as expertise. In law, accounting, and advisory work, the first firm that can understand a messy situation quickly and respond with confidence often wins the mandate. Faster document intake and faster issue framing are therefore not just delivery improvements. They are part of business development.

The economics are unusually clean. Suppose a senior accountant or advisor costs a firm roughly $90 to $150 per billable hour when salary, benefits, overhead, and utilization are taken together. Saving 55 minutes on a recurring document review task can preserve most of an hour of high-value capacity. If that kind of task appears only five times a week, the annual recovered capacity is already meaningful. Add the shorter tax research cycles and the ability to turn around prospects faster, and the software starts behaving less like a convenience tool and more like a margin engine. This is an estimate, not the firm’s disclosed internal model, but the direction is difficult to miss.

There is also a quality effect that small firms should take seriously. When professionals are overloaded, they do not only move slower. They simplify their own process. They read less broadly, defer harder questions, and reserve deep work for only the most obviously risky engagements. AI-supported retrieval changes that pattern because it lowers the cost of being thorough. A team that can scan more source material quickly is more likely to spot a buried clause, an unusual tax treatment, or a missing support document before the issue becomes expensive.

For legal, accounting, and advisory firms, the lesson is narrower and more practical than “adopt AI.” Start where expert time is being wasted on document digestion rather than expert judgment. Contracts, leases, partnership agreements, policy manuals, diligence files, and tax memoranda are good candidates because they are dense, repetitive, and expensive to parse manually. The firms that get the most value will not be the ones that ask AI to replace expertise. They will be the ones that redesign the first draft of professional thinking so their experts spend more of the hour on interpretation, client advice, and risk calls.

Harris, Hardy & Johnstone’s 2025 case is useful because it shows what adoption looks like when AI is tied to a real unit of work. The before state was familiar: slow review, fragmented research, and hidden capacity limits. The after state was not magic. It was a tighter workflow, a faster route to judgment, and a stronger ability to win and deliver work with the same team.