Managed AI for Columbus Accounting, Law and Engineering Firms
It is Thursday afternoon. An engagement letter has to go out, a long agreement has to be read before Monday, and an RFP is sitting there asking questions your firm has already answered in almost these words. If nobody has told your people where the line is, some of that work is already going into a personal chatbot account, and there is no log of it anywhere you can reach. Almost everything a Central Ohio accounting, law or engineering firm sells leaves the building as a document, which is why this work meets AI early and why the confidentiality question has to be answered first.
Where it starts
The recurring work worth handing over first
Not the impressive demo. The jobs that come up every week and eat somebody’s afternoon, which is where the hours actually come back.
Proposals and RFP responses
Your firm answers the same questions about staffing, methodology, insurance, references and conflicts on every submission, and each new RFP starts with somebody digging through last year’s file. The workflow searches your past responses, matches the closest prior answer to each question, and produces a first draft in the format the client asked for, with the source response noted beside each answer so the person finishing it can see what they are editing from.
Engagement letters and scope memos
Engagement letters vary mostly in scope, fee structure and a handful of standard clauses. A template built on your own approved letters produces a draft from a short set of inputs, flags where the scope language departs from your standard, and leaves fees and signature to a person. The template is built on your letters and stays in your workspace, so your language stays your language.
Long documents and contract review
A long agreement, a bid package, a due diligence set, a technical report. The workflow produces a summary with page references and pulls defined terms, dates, obligations, renewal windows and termination triggers into a list, quoting the sentence each item came from. The reviewer works from the quotes back to the source pages instead of hunting for them.
Meeting notes into a matter record
A client call or a site meeting produces a recording. The workflow turns it into a short summary, a list of decisions, and a list of actions with an owner and a date on each, formatted the way your practice management system takes them. What it replaces is somebody’s memory and three lines in a time entry.
Routine client correspondence
First drafts of the correspondence that fills the gaps in a timekeeper’s day: status updates, requests for missing documents, scheduling, acknowledgments, written in the style of the person whose name goes on it. The draft lands in that person’s drafts folder. Nothing sends itself.
New engagement setup
A new engagement produces the same artifacts every time: an information request list, a kickoff agenda, an internal background memo on the client and its sector, and the intake summary your conflicts check runs against in your own system. The workflow drafts that set from the intake record, so week one starts from drafts instead of an empty folder.
How do we stop people pasting client files into a free chatbot without banning the whole thing?
Bans do not hold. People use the tool in front of them to finish the work in front of them, and a firm-wide prohibition moves that activity onto personal accounts and personal phones. There is no log there, no retention setting, and no way to answer a client who asks what happened to their document. Managed AI puts one company workspace in the middle. Everyone signs in through the firm, so work that was already happening becomes work you can see.
Seats are per person, because that is what licensing is. Usage allowance is set by team and department: the tax group gets an allowance, the litigation group gets an allowance, and the firm sets a ceiling on total spend so the bill cannot run past the number you approved. Each month you get a report of what kinds of work AI is being used for across the firm, in categories like research, correspondence drafting and document summarization, so the managing partner can see whether the written policy is holding.
Control is the ceiling on spend. Governance is knowing what the firm is using AI for. Those two contain the risk, and they are what a managing partner has to be able to answer for at a partner meeting. ScaleSight AI runs on a SOC 2 Type II audited platform, and we will walk whoever owns risk at your firm through how data is handled before anything is signed.
Our work is confidential and some of it is privileged. Where does the line get drawn?
The line gets written down before anyone logs in, and your firm writes it. Some categories usually stay off the workspace entirely: matters under a protective order, sealed material, anything a client’s outside counsel guidelines restrict, and any engagement where your engagement letter has not yet said what it says about AI. On the tax side, Internal Revenue Code Section 7216 and its regulations restrict how a preparer may use or disclose tax return information, and consent rules apply to uses outside preparing the return, so that category gets its own line. Where a client has written AI terms into its guidelines or its engagement terms, that restriction goes into the firm policy, and the policy exists before the first login.
Professional liability does not move because a draft came out of a workspace. Whatever leaves the firm carries a name on it and the review that name implies, and on the engineering side it carries a seal. So the workflows are built to make review fast: a summarization or extraction step shows the passage each statement came from, with its page reference, so the reviewer checks against the source instead of taking the output on trust. Anything addressed to a client lands in a draft folder and waits for a person.
Now the boundary that matters most. Anything built on your documents, your data or a process specific to your firm stays inside your workspace. Your engagement letter templates, your workpaper standards, your model clauses, your specification language, your house style: those are built for you and they stay there. They are never copied into the shared library and never appear in another client’s workspace, including a firm across town that competes with you for the same work.
If AI makes us faster, do we not just bill fewer hours?
On hourly work, yes, and it is better to say that now than three months in. ABA Formal Opinion 512, issued in July 2024, is direct about it: a lawyer charging an hourly rate bills the time actually spent, so time a tool saves is time that does not go on the invoice. The same opinion says a lawyer may not bill a client for the general time spent learning to use such a tool. Accounting and engineering practices are not bound by that opinion, but any firm that sells hours faces the same arithmetic.
The return shows up first in the work nobody pays you for. Proposals, RFP and RFQ responses, statements of qualifications, pitch materials, engagement letters, internal memos, marketing, and the setup that runs at the start of every new engagement. None of that is billable today. Every hour taken out of it comes straight back to the firm.
After that come fixed-fee and capped-fee work, where the efficiency stays with the firm, and the write-offs that happen when a matter runs past what you can reasonably bill. Then there is capacity. If you cannot hire the next associate or the next staff accountant, the work you can take on is set by the hours you already have.
One caveat, said plainly. If your firm bills almost everything hourly, holds a comfortable realization rate, and is not turning work away, the commercial case here is thinner, and we will say so on the call before you spend anything. The case is strongest in accounting practices heading into busy season, in engineering and architecture practices doing fixed-fee project work, and in any firm where the proposal and RFP load has quietly become somebody’s second job.
What does this give us that firm-wide ChatGPT licenses do not?
Start with what happens the next time the model changes. Work built straight onto one AI company’s model carries that company’s pricing and release decisions with it, so a repricing or a retired version turns into rebuild work at your expense. Your workspace sits above the models. When one changes, we re-point the work and it keeps running, and the person using it never learns anything happened.
The second reason is that the lead changes by task. One model reasons better over a long agreement. Another handles live web research. Another produces images. Another drafts cheaply enough to run the same summary across a whole document set. Each job is routed to whichever model does it best, and the person typing does not have to keep a scoreboard of which one is ahead this quarter.
Then there is the library. Work that shows up in the same shape at every company is built once and shipped into every client’s workspace: a recording turned into decisions and dated action items, a long PDF reduced to a one-page brief with page cites, a first draft reply written in the sender’s own style. New templates, agents and workflows arrive every month, so a firm that starts in March inherits everything built before March and keeps receiving what comes after. Work built on your own documents, your own data or your own process stays in your workspace and is never shared.
That is where the two halves meet. When somebody in your firm works out a better way to turn deposition transcripts into a chronology, or a cleaner way to draft a submittal response, that method gets written up as a template so nobody else has to invent it again. Collaboration is your method spreading inside the firm. Acceleration is our library growing whether or not you asked for anything this month. In a year, the best AI ideas in your firm should come from your people, not from us.
What does this cost, and how do we keep it from running away?
Pricing is per seat plus usage. A seat is a person who can sign in. Each seat carries a monthly usage allowance, and your firm sets a ceiling on total spend, so the number you approved is the number that can appear on an invoice. A $250 monthly minimum applies.
Budget is allocated by team and department. A person gets a seat. A department gets a budget. If litigation support is running heavy document review in March, you move budget to that department for March without raising the allowance for everyone else in the firm.
Published ranges are monthly totals for the whole firm at a given seat count. Up to 10 seats runs $250 to $500 a month. The 10 to 25 seat band runs $500 to $1,175. The 25 to 100 seat band runs $1,175 to $3,800. Almost everyone in a professional firm works at a desk, so your seat count sits close to your headcount, unlike a plant or a job site where most of the payroll never logs in. Expect the usage half of the bill to move month to month with busy season and proposal cycles.
The ceiling is the part that matters if a software line item has surprised you before. Usage that would take the firm past the number you set does not quietly happen anyway. The monthly report shows usage against each team’s allowance and the kinds of work it went to, so the question at the partner meeting is whether the tax group’s allowance is set correctly.
ScaleSight AI works out of 1733 W Lane Ave in Columbus, an office it shares with TTS Cyber, the managed IT and cybersecurity company in the same Koine Cyber group. Questions about who can reach a system, and how long records are kept, sit with them.
Questions
Asked by professional services businesses
Is ScaleSight AI SOC 2 Type II certified?
ScaleSight AI runs on a SOC 2 Type II audited platform, and ScaleSight AI is not itself SOC 2 Type II certified. The workspace is branded ScaleSight AI and white-labeled from a third-party vendor, and the SOC 2 Type II audit belongs to that vendor rather than to ScaleSight AI as a legal entity. We hold to that wording because your own clients will ask you the same question, and a firm that passes along a certification claim it cannot support has a problem we do not want to create.
Can we keep certain clients or matters off it completely?
Yes, entire practice groups, specific clients and individual matters can stay off a ScaleSight AI workspace, and it is a reasonable way to start. Seats are granted per person and usage allowance is set by team, so a group that is not on the workspace simply has no seats. Where a client’s outside counsel guidelines or your own engagement letter restrict AI use on a matter, that restriction goes into your firm’s written policy before anyone logs in, and the workspace does not override the policy.
We could buy everyone a ChatGPT or Copilot license. What does a workspace add?
Three things a stack of individual licenses cannot give you. Work built directly on one AI company’s model has to be rebuilt when that company reprices or retires it, while work in a ScaleSight AI workspace gets re-pointed and keeps running. Each job runs on whichever model handles it best, so reasoning over a long agreement and cheap bulk drafting do not have to happen in the same place. And the firm gets a spend ceiling, budget set by team and department, a monthly report of what kinds of work AI is being used for, and new templates every month from the shared library, with anything built on your own documents or process staying in your workspace.
What would a 40-person firm actually pay per month?
A 40-person firm sits in the 25 to 100 seat band, which runs $1,175 to $3,800 per month as a total for the firm. Pricing is per seat plus usage: each seat carries a monthly usage allowance, your firm sets the ceiling on total spend, and a $250 monthly minimum applies. Because nearly everyone in a professional firm works at a desk, your seat count tracks your headcount closely, so plan on that band instead of the one below it. The usage half of the bill moves with workload, which for most firms means busy season and proposal cycles.
When would we need an AI Build Project?
When the job depends on your own documents or your own process, which is the work a shared library cannot cover. A proposal assembler that reads your firm’s past RFP responses is one. So is an intake workflow shaped around your practice management system. The scope and the end date are agreed before the work starts, and what gets built lands in your workspace and stays there. Managed AI is the ongoing service that runs it afterward, keeps it working when models change, and reports on how it is being used.
Other industries
We also work with
Tell us what your week actually looks like.
The readiness review starts by finding where AI is already being used in your company, what it is costing, and which recurring work is worth handing over first. You keep the findings either way.