Every agency with a workload problem is choosing between three options, whether or not anyone has named them. You can build something, you can hire someone, or you can run the work on a system that already exists.
Most agencies only seriously consider two. Hiring is familiar, building sounds modern, and the third is harder to picture because it does not map onto a job posting or a software purchase. That is usually the one that gets skipped, and sometimes it is the right answer.
This piece walks the three honestly, including what each is genuinely good at, and then covers the two questions that decide between them: where the licensing line sits, and what actually happens to the hours you free up.
Before choosing, find out where the hours go
The choice is unanswerable without this, and almost nobody has the number to hand.
Not headcount, and not a feeling that the team is stretched. A count of where the service week actually goes, by request type. Most owners are surprised twice: by how much of the week goes to a small number of repetitive request types, and by how much of that lands on their most expensive people. Until that is written down, every option is being evaluated against a guess. It is worth auditing where the service week goes before doing anything else.
The three options, honestly
Build
Build means creating your own automation, whether that is a developer, a no-code stack, or an ambitious ops manager with a lot of integrations.
What it is genuinely good for: the workflows that are specific to your agency and that no vendor will ever prioritise. Nobody understands your carrier mix and your exceptions better than you do.
What it costs, beyond the build: maintenance forever, and the fact that automation needs standardized work underneath it. If a request is handled three different ways by three different people, automating it just picks one of the three and makes the inconsistency faster. That is the most common reason a promising internal project quietly stops, and it is the same reason AI tools alone do not fix an agency.
Hire
Hiring is the default because it is legible. You know how to write the posting and how to budget the seat.
What it is genuinely good for: work that needs judgment, relationship, and context, and volume that has settled at a permanently higher level rather than spiking.
What it costs: the seat is permanent while the volume may not be, and the lead time is long enough that a hire made for this quarter arrives for the next one. It also does not change the operating model. A second person doing undocumented work the same way produces two versions of it, which is how the drift starts.
Run it on a system
The third option is that the work is defined as tasks, routed by license and skill, with the routine execution handled by software and the judgment kept with your people.
What it is genuinely good for: high-volume, well-defined request types where the cost is real and the judgment is low. Certificates, endorsement processing, and renewal administration are the usual candidates, which is why starting with certificates is the common first move.
What it costs: the standardization has to happen. This is the same prerequisite the build option has, which is worth sitting with, because it means the work of defining your processes is not wasted whichever way you go.
The question that decides it: where is the line?
Every owner arrives at the same question. If work moves off my team, where exactly does it stop, and what can be handled by software or by someone unlicensed?
The answer is not a category, it is a step. Service work decomposes into steps with different requirements, and only some of them legally require a licensed person. A renewal is not one job: the remarketing conversation needs a licensed person, and pulling the expiring policy does not. Drawing that line at the request level rather than the step level is what makes people conclude the whole thing is impossible.
The line has to be enforced rather than remembered, which is what routing by license and skill means, and it is the mechanism behind the licensing bottleneck easing. If AI is doing any of it, the same applies with a confidence threshold and an audit trail, so the work stays explainable to a regulator or a buyer.
Then what happens to the hours?
This is the question that separates a workload fix from a growth decision, and it is the one most often skipped.
Freed hours do not automatically become anything. Left alone they get reabsorbed by whatever was being deferred, and six months later the team is as busy as before with nothing obvious to show for it. That is not a failure of the change, it is the absence of a decision about what the time is for.
The agencies that get something out of it decide in advance: this many hours go to renewals and rounding, this many to new business, this many to the accounts nobody has called in a year. Deciding afterwards is deciding never.
S&G Mitchell went from 17.9% to 60%+ EBITDA in 12 months. Same agency, same customers, same book of business. What changed was the operating model underneath.
What the production data shows
Across the agencies we operate:
- AI triage filters about 68% of inbound as noise, meaning carrier notifications, duplicates, and automated confirmations that never needed a person. Classification runs at a 98% triage success rate, in under 10 seconds per item.
- About 40,000 tasks a week are classified and routed with no manual intervention.
- Roughly 44% of completed service tasks are executed by AI, with the rest routed to people by license and skill.
- Median handle time on a task is 3.8 minutes, down 16% from 4.5.
- Escalations, meaning a task where someone is stuck and has to hand it on, are down 20%, from 435 a day to 348, and still declining.
COVU runs on top of your AMS, not instead of it, with 8 AMS integrations live. Your own staff stays the default, and this is extra capacity when you need it.
Get the longer version
This piece is the short form of an argument that takes more room than a blog post. The AI-Powered Agency is a free 25 page guide for independent P&C agencies covering the same five questions in detail: where the hours actually go, where the line sits when work moves off your team, how you know a request is getting done without chasing it, what comes off the team first and in what order, and what to do with the capacity once you have it.
Download the guide if you are weighing these three options and want the sequencing rather than the summary.
Frequently Asked Questions
Should an insurance agency build its own automation or buy it?
Build is strongest for workflows specific to your agency that no vendor will prioritise, and weakest as a general answer, because automation needs standardized work underneath it and then needs maintaining forever. Running on an existing system is stronger for high-volume, well-defined request types. Both have the same prerequisite, so defining your processes is useful either way.
How do we know whether to hire or add capacity another way?
Look at whether the volume is a peak or a permanent step up. Volume that settles at a new, higher floor justifies a hire. Volume that returns to baseline does not, and the hire usually arrives after the pressure has passed. Staffing for your average rather than your peak covers how to tell them apart.
What work can move off a licensed team member?
The steps within a request that do not legally require a license, which is a step-level question rather than a category-level one. The boundary has to be enforced by the system rather than remembered by whoever picks the work up. Requirements vary by state and by carrier agreement, so confirm the specifics with your own counsel or compliance lead.
What is in The AI-Powered Agency guide?
It is a free 25 page guide for independent P&C agencies covering five questions: where the hours go, where the line sits when work moves off your team, how you know work is getting done without chasing it, what to offload first and in what order, and what happens to the freed capacity. You can download it here.
Will freeing up hours actually grow the agency?
Only if the hours are assigned in advance. Left undirected they get reabsorbed by deferred work and the team ends up as busy as before. Deciding beforehand what the time is for, whether that is renewals, rounding, or new business, is what turns a workload fix into a growth outcome.