# The One Workflow We'd Automate First in Your Industry

> Contractors, medical practices, law firms, accountants, manufacturers, property managers. Six industries, six different piles of paperwork, and where each one should actually begin.

- Canonical: https://www.rnits.com/blog/ai-automation-by-industry-where-to-start
- Provider: RNITS — cybersecurity-first managed IT for small businesses
- Phone: (978) 226-8931
- Talk to RNITS: https://www.rnits.com/contact · Free cyber security audit: https://www.rnits.com/free-cyber-security-audit
- Published: 2026-08-01
- Category: Business Solutions
- Tags: AI Automation, Workflow Automation, Small Business IT, Productivity

Last time we went through [the five places automation pays in almost any business](https://www.rnits.com/blog/ai-business-process-automation-small-business) — proposals, repetitive forms, email, intake, and the marketing upkeep that quietly stopped happening. All five apply whether you pour concrete or file motions.

The specifics do not. Which form, which system, which regulator is watching, and which mistake costs you a client versus a license — those are entirely different, and they decide where you should start.

So this is the promised follow-up: six industries we work in across New Hampshire and Massachusetts, what the paperwork actually is in each one, what the credible data says about it, and the single workflow we would build first if you handed us the keys.

One warning about that last part. In every one of these, the first workflow is not the famous one. It is rarely the task that generates the most complaining, and it is almost never the one a vendor demoed to you. The first workflow is the one that is high-volume, low-judgment, and cheap to check — because that is the one that earns enough trust to make the second one possible.

A note on the numbers below. Some come from real surveys with published methodology. Some come from companies selling the software that fixes the problem they measured, which does not make them false but does make them convenient. We have flagged which is which, because you should weigh them differently.

## General contractors and the trades

The paperwork is prequalification packets, insurance certificates, bid packages, RFIs, submittals, change orders, and lien waivers. On public work in Massachusetts, add the prequalification and filed sub-bid paperwork that shows up before you have earned a dollar.

RFIs are the famous number here. The Navigant Construction Forum's study put the average cost to review and respond to a single RFI at roughly $1,080, with a median of about 9.7 days from creation to closure. That study is over a decade old and its project sample skews large — 800 RFIs on a job is not your world if you run a twenty-person shop. Do not import the headline. Do take the shape of it seriously: an RFI is expensive because it stalls work while it sits, not because writing it is hard.

Which is exactly why RFIs are the wrong place to start. The delay is usually the architect's or the owner's, not yours, and automating your half of a conversation you do not control buys you very little.

**What we would build first: bid package and prequalification assembly.**

Watch what happens when a bid invitation lands. Someone pulls the last similar proposal. Someone hunts for the current certificate of insurance, the bonding letter, the safety record, the list of comparable projects, the W-9, and the licenses — all of which exist, all of which are current, and none of which are in one place. Someone rewrites the company background paragraph again. Then somebody reads the instructions to confirm every required attachment is actually attached, which is the step that gets skipped at 6pm on a due date.

None of that is estimating. Estimating is scoping the work and pricing it, and that stays with your estimator. The assembly around it is mechanical, it is identical across every bid, and it is the reason your best estimator spends evenings formatting instead of takeoffs.

Build a workflow that keeps your qualification documents current in one place, assembles the standard packet, and checks the submitted package against the invitation's requirement list. The payoff is not hours saved. It is that you bid more jobs, and you stop losing one a year to a missing attachment.

## Medical and dental practices

The paperwork is patient intake, insurance eligibility, prior authorization, referrals, and claim denials. This is the industry where the administrative burden is not merely expensive — it delays care, and the people doing it know that.

The data here is unusually good, because the AMA measures it every year. Its 2025 physician survey, fielded to 1,000 practicing physicians, found practices complete an average of 40 prior authorization requests per physician per week and spend about 13 hours of physician and staff time on them. Two in five practices employ someone whose entire job is prior authorization. Ninety-four percent of physicians said it contributes to burnout, and more than one in four reported that the process had led to a serious adverse event for one of their patients.

CAQH, the industry association that indexes administrative transactions, puts a manual prior authorization at roughly 24 minutes of provider staff time by phone or fax, about 16 minutes through a payer portal, and around $3.41 per transaction against about a nickel when it runs fully electronically. Treat CAQH's cost figures as directionally right rather than precise — they lean on self-reported data from the people being measured.

Here is where we will disappoint you. We would not start with prior authorization.

Not because it is not the biggest number. It is. But prior auth runs through payer portals that fight automation on purpose, the rules change per plan, the stakes are clinical, and a workflow that gets it subtly wrong produces a denied claim and a delayed patient. That is the opposite of a cheap-to-check task. It is a fine second or third project, once the practice has seen the approach work.

**What we would build first: eligibility and benefit verification before the visit, plus a prior authorization status board.**

Verification is high-volume, rule-based, and instantly checkable — the answer is either right or it is not, and you find out before anyone is in a chair. Getting it done ahead of the appointment is also what prevents the downstream denial, which means you are attacking the prior-auth problem from the end where mistakes are cheap.

The status board is the unglamorous half and the one your staff will thank you for. Most practices cannot answer "where does this authorization stand" without someone opening four portals. A workflow that tracks every outstanding request, its age, and who touched it last does not submit anything — it just makes the queue visible. That alone stops the ones that fall through.

Everything in this category is PHI, which changes the engineering. You need a business associate agreement with whatever platform you use, the consumer tiers of these tools are not eligible, and access has to be scoped so a workflow can read what it needs and nothing else. That is the same discipline that lives in our [HIPAA compliance work](https://www.rnits.com/services/hipaa-compliance-services), and it is not optional here.

## Law firms

The paperwork is intake, conflict checks, engagement letters, discovery, document assembly, and time entry.

Clio's Legal Trends research puts the average utilization rate at about 38% — roughly 3.0 billable hours captured in an eight-hour day. Apply an 88% realization rate and about 2.6 of those hours actually get invoiced. Clio sells practice management software, so read the framing with that in mind, but the underlying finding has been stable for years across sources: most of a lawyer's day is not billable, and most of the non-billable part is administrative rather than business development.

The temptation is to point AI at drafting. Resist it for now. Drafting is where the judgment is, where the malpractice exposure is, and where a confident wrong answer looks exactly like a right one.

**What we would build first: intake capture and first response.**

A prospective client calls, fills out a web form, or emails an address three people half-monitor. The response time on that first contact is the single biggest predictor of whether they retain you, and it is almost entirely a routing problem. Nothing about it requires legal judgment — it requires that the message reach a human quickly, with the relevant facts already pulled together.

Build the workflow that captures every inbound matter from every channel into one queue, classifies the practice area, runs the preliminary conflict search against your existing client list, and drafts the acknowledgment for a person to send. Note the word preliminary. The conflict check is surfaced for a human to clear, never cleared automatically. That is the line, and it does not move.

The second thing worth building is contemporaneous time capture, because the gap between 3.0 hours worked and 2.6 hours invoiced is partly reconstruction at the end of the week.

Confidentiality changes the rules here the same way PHI does in a practice. Privileged material does not go into a personal AI account, and the tenant it does live in needs to be configured deliberately — which is the [Microsoft 365 management](https://www.rnits.com/services/microsoft-365-managed-services) side of the work rather than an afterthought.

## Accounting and bookkeeping firms

The paperwork is client onboarding, engagement letters, the document chase, and the reconciliation of whatever the client sent against what you actually asked for.

This is the industry where the public numbers are worst, and we would rather say so than dress it up. Almost every statistic about "hours spent chasing client documents" traces back to a company selling a document portal, and the figures are suspiciously round. Thomson Reuters' 2025 State of Tax Professionals report is the more credible read, and it is less quotable — efficiency and growth top the strategic agenda, about 37% of tax teams are automating compliance processes, and talent shortages are the binding constraint. Nobody has published a trustworthy number for how many hours busy season loses to unanswered document requests, so we will not invent one. Every firm we have asked describes it as one of the top two problems and none of them measure it, which is itself the finding.

**What we would build first: the document request cycle.**

One list per client of what you need, generated from what that client's engagement actually requires rather than a static checklist. Automatic reminders on a schedule, so nobody on your staff has to decide whether it is rude to follow up a third time. A running status anyone in the firm can see without asking. And an intake step that checks what arrived against what was requested, so the missing 1099 gets flagged the day it did not show up instead of three weeks later when a preparer opens the file.

The reason this is first, rather than the more glamorous return-preparation automation, is that it compounds. It applies to every client, it runs the whole season, it requires no judgment, and every mistake it makes is visible immediately. It is also the workflow that most directly buys back the thing you cannot hire your way out of — capacity in March.

Take the baseline before you build it. Count how many request rounds a typical client takes today. Without that number you will spend next April arguing about whether it helped.

## Manufacturers and machine shops

The paperwork is RFQs, purchase orders, order acknowledgments, routers and travelers, material certifications, and — for anyone in the defense supply chain — the compliance documentation that decides whether you stay on the approved list.

Fictiv's 2025 State of Manufacturing report puts the average complex RFQ at about 11.5 hours of work, with roughly 18% of that time spent on anything you would call value-adding. The rest is looking for data, waiting on approvals, and reconciling systems that do not talk. Fictiv is a manufacturing marketplace, so that report is not disinterested. It also matches what we see on shop floors closely enough that we will cite it.

On the demand side, buyer surveys consistently find most procurement people expect a quote inside 24 hours and very few will wait past three days, while typical shop turnaround is one to three days. Those surveys are run by quoting-software vendors, so discount them — but every shop owner we know has lost a job to a faster competitor and can name it.

**What we would build first: RFQ triage.**

Triage rather than quote generation, because most shops quote too many jobs and the estimator's time is the scarcest resource in the building.

The workflow reads inbound RFQs — from email, from portals, from the PDFs and BOMs that arrive as attachments — extracts the part, quantity, material, tolerance, and required date, and scores each one against what you actually want to run: is this in your capability envelope, is the quantity in your sweet spot, is this a customer who buys or a customer who collects quotes? Then it routes the worthwhile ones to the estimator with the data already pulled, and flags the rest for a fast, polite decline.

That last part is worth more than it sounds. Declining in an hour preserves the relationship. Silence for four days does not.

Quote assembly is the natural second build, once triage has proven the extraction is accurate. And if you supply defense primes, the certification and compliance paperwork is its own project — the requirements are specific enough that it belongs in a [CMMC compliance](https://www.rnits.com/services/cmmc-compliance-services) conversation rather than a general automation one.

## Property managers

The paperwork is work orders, vendor coordination, tenant communication, lease renewals, owner reporting, and the applications and screening that come with turnover.

Buildium and NARPM's 2026 industry report found 38% of rental owners name maintenance as their top source of stress — ahead of vacancies, residents, and rent collection — and it ranks as the second-highest operational challenge for property managers. Hemlane, looking at more than 193,000 maintenance requests in its own system, reports a median resolution of about 13 days. Read that as one company's book of business rather than an industry census.

The more useful finding, and the one that shapes what to build, is that tenant satisfaction tracks communication quality more closely than it tracks actual repair speed. A resident who knows a part is on order Thursday is calmer than one whose faster repair happened in silence.

**What we would build first: work order intake and automatic status updates.**

Requests arrive by text, by portal, by phone, and by a resident stopping someone in the parking lot. Consolidating them into one queue, classifying by urgency and trade, and routing to the right vendor is the ordinary half of the build.

The part that changes your week is the status updates. Every stage change — received, assigned, scheduled, parts ordered, completed — generates a message to the resident automatically. That is not a technology achievement. It is the difference between a resident who waits and a resident who calls you three times, then complains to the owner, then posts a review. Most of the phone calls your office fields are not new information requests. They are people who have no idea what is happening.

Vendor dispatch decisions and anything involving money stay with a person. The workflow assembles and informs. It does not spend.

## If your industry is not on this list

The pattern generalizes. Look for the task where somebody retypes information that already exists in a system you already own, where the rules are consistent, where a mistake is visible immediately, and where the person doing it would be relieved to stop. That is your first workflow, whatever business you are in.

Then check it against the same three questions from the last post before anyone configures anything. What is the baseline — how many of these a week, how long does one take now? What happens to the weird one, and does the workflow know when it is out of its depth? Where is the approval gate, and is it on every step where money leaves or a client hears from you?

Notice what is absent from all six recommendations above. Not one of them is a workflow that talks to your customers unsupervised, and not one of them makes a decision that costs money. Every one assembles, sorts, tracks, or checks, and then a person acts. We pick them for that reason. Those are the projects that survive contact with a real office, and the ambitious ones account for most of the failures.

## Where we come in

We are an IT and cybersecurity company in Tyngsboro, Massachusetts, working with small businesses across New Hampshire and Massachusetts — onsite within about 150 miles, remote nationally. We are not an AI consultancy and there is no five-figure discovery phase attached to any of this.

What we do is the missing part: spend a few hours with the people who actually do the work, map the process including the exceptions everybody handles by instinct, and identify the one or two workflows that pay for themselves inside a quarter. Then design them, connect them to the systems you already run, put approval gates where money and client data are involved, and [train the people who have to live with them](https://www.rnits.com/services/ai-training). The design and build side of that is our [AI automation](https://www.rnits.com/services/ai-automation) practice.

We stay stubborn about permissions. When a workflow connects to QuickBooks, a practice management system, or your Microsoft 365 tenant, it inherits whatever access you hand it, and the fast path is always to hand over everything. We do the slower version — narrowest access that still finishes the job — which matters more in a regulated industry than anywhere else, because the audit question is not whether the automation worked but who could see what.

If you want a straight conversation about which part of your week is worth automating, [get in touch](https://www.rnits.com/contact). We will walk your real processes with you and tell you where the hours are hiding, including the cases where the honest answer is to fix the process or drop it rather than automate it.

The Rnits Company. The un-MSP. (978) 226-8931.

---

**RNITS** — New Hampshire & Massachusetts (onsite within 150 miles of Tyngsboro, MA); remote support nationwide

RNITS (The Rnits Company) · 404 Middlesex Road, Suite 9, Tyngsboro, MA 01879 · (978) 226-8931 · info@rnits.com

Contact: https://www.rnits.com/contact · Free cyber security audit: https://www.rnits.com/free-cyber-security-audit