For leaders around the Capital Region who've heard "you should automate that" a dozen times and still don't know where to begin.
Go find the thing your team spend more time with. Usually it's a weekly report, a filing, or that hand off where Sam in billing retypes numbers from one screen into another screen. If it comes around on a schedule, the data is already sitting somewhere, a typo causes headaches, and there's more of it every time you grow, congratulations. You've found your first automation opportunity.
It is not either the flashy one or the chatbot on the website. It is the boring one.
Picture Thursday afternoon at your office. Somebody, probably someone you'd hate to lose, opens the same spreadsheet they opened last Thursday. Hours from here, prices from there, paste, fix the columns that shifted again, again add the totals by hand because of that one time in March. Send. Nobody loves this part of the week, the least of all, and somehow everybody's decided it just comes with the job. But it doesn't, really.
Why do most companies pick the wrong first automation?
Because they start from the tool instead of the work.
Somebody sees a demo, gets excited about AI agents, and the first project becomes "let's build something smart." Six months later there's an agent nobody uses, and the Thursday spreadsheet is still sitting there.
We see the same handful of mistakes over and over with Albany companies:
- Picking the impressive project. A customer facing AI assistant sounds great in a board meeting. It's also the hardest thing to get right, and the easiest to quietly abandon.
- Asking leadership what to automate. Leaders know the business. The people doing the work know where the hours actually go. Those are two different lists.
- Automating a mess. Got a seven step process that's only seven steps because "that's how Linda set it up in 2017"? Automate it as is and congrats, you now have seven steps that break faster.
- Forgetting who keeps it running. Systems change. A vendor updates a portal, a column moves, a password expires. An automation nobody maintains breaks on a random Tuesday.
So what fixes all four? Go small. Pick something real, something boring, and something where you'll be able to say "this used to take six hours and now it takes forty minutes."
Which processes should you automate first?
Run every candidate past four quick questions. We ask them in pretty much every first meeting, and honestly they hold up whether we're talking to a 40 person CPA firm or a 250 person manufacturer.
Four checks for your first automation
- Does it come around on a schedule? Fridays, month end close, every new invoice, every new hire. Something that happened once last spring doesn't count.
- Is the data already sitting somewhere? A spreadsheet, QuickBooks, your CRM, a vendor portal. Somebody's job is just moving it from there to here.
- Does a mistake hurt? Rework. A filing that comes back rejected. A late payment. That phone call to a client nobody volunteers for.
- Does it get bigger when you do? Land ten new clients and you've got ten more rounds of the same chore. Keep going and eventually you're posting a job just to keep up with it.
Three yeses, take a closer look. Four, and you're likely paying for it already in somebody's week.
Where we tend to find these:
- Reporting and compliance. Month end reports, grant paperwork, claims, anything a regulator wants on a deadline.
- Finance and admin. Two systems that never quite agree, invoices piling up in an inbox, chasing people for documents.
- Onboarding. Creating accounts, collecting forms, sending the same five emails to every new employee or client.
- Data between systems. Your CRM has one number, accounting has another, and guess who gets to figure out which one's right.
How do you know if a process is costing you enough to automate?
Do the napkin math before anyone builds anything. It takes five minutes and it settles a lot of arguments.
Hours spent per week, times what that person's hour really costs you (salary plus benefits and overhead), times 52.
Say someone spends six hours a week on a report and their time costs you about $40 an hour. That's roughly $12,480 a year, for one report, from one person. Now add the time somebody else spends checking it, and whatever it costs when it goes out wrong.
Don't get hung up on the exact figure, though. It's the comparison that tells you something. Low cost and a process that keeps changing on you? Let it go. Real money and a process that's looked the same for two years? Now we're talking.
And there's a line the napkin usually leaves off: what that person would be doing with those hours back. Calling clients. Fixing the next thing. Going home at five.
What does a good first automation look like in practice?
Here's a real one. A New York solar and wind contractor relies on NYSERDA reimbursement to fund its projects, and every claim has to show who worked, how many hours, on which project, and at what location. Each location also has to be checked against Disadvantaged Community status.
The hours already lived in the engineers' spreadsheets. But someone on staff was spending more than a full day every week copying them by hand into NYSERDA's format and matching each one to its DAC status.
That checks all four boxes. Weekly. Data already exists. Errors mean rework before claims go out. And every new project adds more of it.
TAG built automation around the spreadsheets the engineers were already using, so nobody changed how they log hours. Now the data flows into the right format on its own, gets checked, and staff just review the exceptions before submitting. The company eliminated about $60,000 a year in manual work, cut report prep time by more than 70%, and hasn't had to add admin staff as the project load grows. You can read the full NYSERDA reporting automation case study.
Notice what's missing from that story. No chatbot. No big platform rollout. One annoying weekly chore, handled.
Do you need AI to automate a process?
A lot of the time? Nope. Folks tend to raise an eyebrow at that one.
Look at where the hours go. It's mostly stuff that happens exactly the same way every time, same steps, same order. Plain old automation is great at that. It runs cheaper than an AI model, it's quicker, and it never makes things up.
AI earns its place on the steps no rule can cover. Reading an unstructured document. Sorting emails by what they're about. Flagging the record that looks off.
Plain automation vs AI: which one fits the step
| The step looks like | Better fit | Example |
|---|---|---|
| Same steps, same order, every time | Plain automation | Moving hours from a spreadsheet into a reporting format |
| Data has to be read or interpreted | AI | Pulling the key details out of scanned invoices or contracts |
| Mostly routine, with occasional oddballs | Both | Automation does the routine; AI flags exceptions for a person |
| A mistake would be costly or public | Automation plus a human check | Anything going to a regulator, a client or a bank |
Most of the workflows we build use both. Some use no AI at all. If someone pitches you AI for a job a simple rule could handle, ask why.
When should you NOT automate a process yet?
Vendors won't love this section.
If the process changes every month, please just wait. You'd spend your savings just keeping the automation up to date. Ask three people how it works and get three answers? Wait, and get them in a room first. Twenty minutes a month? Leave it. We'd rather not sell you a build you don't need.
And a 30 person shop where the owner still eyeballs every invoice? Might not need any of this yet. The fix there could be a setting buried in software you already pay for. A shared template. Or finally cleaning up the spreadsheet everybody's scared to open.
Automate it, fix it first, or leave it alone
| What you see | What to do |
|---|---|
| Stable, recurring, data already exists, errors hurt | Automate it |
| Recurring, but everyone does it differently | Fix the process first, then automate |
| Rare, small or changing constantly | Leave it alone for now |
Honest answers here save money. A good partner will tell you when the answer is "not yet."
Why the first automation matters more than the biggest one
Whatever you do first, people remember.
If it's small, visible and clearly saves time, people start bringing you ideas. The office manager mentions the onboarding emails. Someone in finance asks about the reconciliation. Before long there's a little list going, and every item on it has a dollar figure next to it.
Go big and vague instead, let it run late, and the verdict around the office becomes "eh, automation doesn't work here." Meanwhile that Thursday spreadsheet gets to live another five years.
So think of the first one less as a technology project and more as proof. Proof that this works in your company, with your people, on your systems. That's what makes the second and third ones easy.
How TAG Solutions Approaches Process Automation
TAG Solutions has worked with organizations around Albany and the Capital Region for more than 30 years. Every client has an assigned vCIO, and these days automation comes up in most of their planning conversations.
First thing we do is talk to whoever's actually doing the work. Leadership too, sure, but the person running the Thursday spreadsheet usually knows more. Out of that comes a list. We put a rough value and effort on each item, and we'll flat out tell you which ones to skip. You get a number before anybody writes a line of code.
After that we build on top of what you already use and test it side by side with the old way. You see it working before any monthly service kicks in. Once it's proven, it moves onto our Automation as a Service plan. We host it and babysit it. Vendor redesigns their portal on a Wednesday? Our headache.
We back everything with service guarantees.
Not sure which of your processes passes the four checks?
It's a quick self-check. Go through it with your team and you'll come out with a short list of the chores eating the most time, plus a pretty good idea of which one to tackle first.
Questions Operations Leaders Ask Us About Automation
How long does it take to automate a business process?
Weeks, typically. Not months. A focused process like a recurring report or a data hand-off between two systems usually goes from mapping to a working version in a few short sprints. The mapping step takes longer than people expect, because that's where you find the exceptions nobody wrote down. Skipping it is how projects drag. Big multi-department rollouts take longer, which is one more reason not to start there.
Is automation a good fit for a company with fewer than 100 employees?
Usually, yes. The catch is picking the right process. In a smaller company the busywork tends to pile up on the same two or three people, so they feel it more than anybody. If one of them burns a full day a week on something that passes all four checks, that's a fifth of their job. It doesn't pay off on work that's rare or keeps shifting. Grab a napkin first: hours a week, times what an hour costs you, times 52.
Will automation replace people on our team?
In the companies we work with, it mostly replaces tasks, not people. The person who used to spend Thursday afternoon copying numbers spends it reviewing exceptions and doing work that needs judgment. Where automation does change hiring is growth: you can take on more clients or projects without adding admin headcount for every jump in volume. Be upfront with your team about that from day one. It makes them a lot more likely to point out the next process to fix.
What does it cost to automate a process?
It depends on the process, so be wary of anyone quoting before they've seen it. The useful comparison is the yearly cost of doing it by hand. Our Automation as a Service model is a small build fee plus a flat monthly fee for hosting, monitoring and maintenance, and the automation has to prove it works before the monthly service starts. That way you're never paying every month for something that didn't deliver.
Who maintains the automation once it's built?
Somebody has to, and it's the part everyone forgets. Portals get redesigned. A file format changes. Passwords expire. Leave an automation alone long enough and one morning it just stops. If you've got someone in-house with the skills and the spare hours, great. Most mid-sized companies we meet don't, so we handle it as a monthly service: we watch it, fix it when it breaks, and tweak it when your systems change.
Not ready for that yet? Get Weekly AI Tips for short, practical ways to put AI and automation to work.


