How do I know if this is a process problem or a people problem?

Watch what happens when a good employee runs the workflow. If it still takes hours, or the same errors appear, it is a process problem. If only one person can do it and everyone else struggles, it is a people or training problem. Automation solves the first. It does not solve the second.

Can we fix this with our existing software, or do we need something built?

Usually both. Most ERPs have configuration options that are not being used, and those should be exhausted first. When the tool genuinely does not exist inside your software, that is when [custom AI solutions](https://throughlinebuilt.com/services/custom-ai-development/custom-ai-solutions/) make sense. An audit is how you find out which category you are in.

What is the difference between automation, AI, and hiring another admin?

Hiring an admin scales linearly with cost. Automation scales with volume once it is built. AI, in the sense we use it, is one component inside an automation: the part that reads unstructured documents or classifies ambiguous cases. The system around it is still deterministic software with human approval.

How much does this cost?

Every engagement starts with a fixed-fee AI Capability Audit at $7,500. The build is a separate fixed fee, quoted in writing after the audit. Cloud costs are metered like any other cloud service and quoted in writing before the build starts. We do not promise a specific ROI figure because we do not know your numbers yet.

When is it too early to automate?

When you cannot describe the chore in one sentence, when the volume is low, or when the process itself is still being figured out. Wait until the workflow is stable and painful. That is the right moment.

Signs Your Business Needs Process Automation (Honestly)

What are the real signs your business needs process automation? The short answer: a person is re-keying data from one system into another, a document sits in an inbox waiting for someone to read it, and a margin leak keeps showing up in month-end that nobody has time to chase. Most posts on this topic are written by software sellers, so every symptom becomes a buying signal. That is not honest. Some of these signs mean you need a system built into your ERP. Some mean you have a process problem no tool will fix. Some mean you should wait.

Key takeaways

If a person copies data from a PDF or email into your ERP every week, that is document work a machine can read and a person can approve.
Rebate and margin leaks usually show up in month-end reconciliation, not in real time; that gap is where money quietly disappears.
A chatbot is not automation. Automation is a system that reads real documents and writes to real records with human approval on anything that touches money.
Throughline's engagement starts with a fixed-fee AI Capability Audit at $7,500; the initial call is free.
If your process is broken on paper, automating it will only make the mess faster. Fix the process first.

Sign 1: Someone is re-keying data every week

If a person on your team opens a PDF, an email, or a portal, reads a number, and types that number into your ERP, you have a candidate for automation. This is the most common pattern we see. It is also the one operators most often accept as normal because "that's just how the vendor sends it."

The fix is not a macro. It is intelligent document automation that reads the document, extracts the fields, and stages an entry for a person to approve before it posts.

Sign 2: Month-end keeps surfacing money you already earned

Rebates, tiered pricing, freight allowances, and volume discounts are all revenue you have technically earned but have to fight to collect. If your finance team is reconciling these after the fact, in spreadsheets, that is not a finance problem. It is a systems problem.

Kelsan, a multi-state distributor in our own group running Epicor P21, had this exact pattern. We built the rebate-and-margin tool inside Kelsan that reads supplier rebate documents, matches them against invoices in P21, and surfaces entries for a person on the team to approve before anything posts. It has recovered over $100,000 in margin that would otherwise have been lost. The mechanism is boring on purpose: a tool reads the documents you already receive, writes entries a person approves, and stops there.

Sign 3: You are hiring an admin to survive a workflow

If the honest reason for the next hire is "we need someone to keep up with the paperwork," that is a signal worth pausing on. Hiring an admin to run a broken workflow is a valid decision. Sometimes it is the right one. But the calculation should be explicit: what does the workflow cost per year in labor, and what would it cost to build a system that reads and writes into your existing software.

That is a real conversation about back office automation, not a pitch.

Sign 4: Your ERP has the data, but nobody trusts the reports

You paid for Epicor, NetSuite, Sage, or Dynamics. The data is in there. But your operators run their real decisions off a spreadsheet that a manager updates on Fridays. That gap between the ERP and the spreadsheet is where automation lives. Not by replacing the ERP. By wiring something on top of it that reads the records and does the boring work the ERP was never configured to do.

This is what ERP automation actually means in practice. It is not a migration. It is a system built into what you already run.

Sign 5: The same exception keeps happening

Every business has exceptions. The question is whether the same exception keeps happening. A pricing override that shows up on the same three customers. A freight charge that gets manually adjusted on the same shipping lane. A credit memo issued for the same reason twice a month.

Recurring exceptions are the clearest signal that a rule can be encoded. Not "AI decides." A rule reads the record, flags the case, and a person on your team approves the adjustment.

Sign 6: You are being told to "use AI" and nobody knows what that means

If the mandate from above is "figure out AI" and the response from below has been a ChatGPT subscription and a Zapier account, you are not doing automation. You are doing AI usage. Those are different problems.

Real business process automation is a system that reads the documents your business already receives, writes entries into the software your business already runs, and hands off to a person for approval on anything that changes a record. It is not a chatbot in the corner of a screen.

Sign 7: Growth is exposing the seams

You grew. Now the workflows that ran fine at half the volume are starting to crack. Orders take longer to enter. Errors compound. Someone stayed late three times last week. Growth is the most common trigger for a real automation conversation because it turns latent inefficiency into visible pain.

When not to automate

Three situations where we would tell you to wait.

Your process is broken on paper. If the workflow itself is unclear, or three people on the same team do it three different ways, a tool will not fix that. Automating a broken process only makes the mess faster. Fix the process, then automate.

The volume is not there yet. If a task happens six times a year, a well-written checklist is cheaper than a build. Automation makes sense when the same work happens often enough that the cost of the build is recovered in a reasonable window.

You have not defined the outcome. "We want to use AI" is not a project. "We want to stop losing rebate dollars because we cannot reconcile supplier statements against P21 invoices" is a project. If you cannot state the chore and the outcome in one sentence, an audit is a better first step than a build.

What a real fix actually looks like

A system, not a subscription. It reads the documents you already receive (PDFs, emails, portal exports). It matches them against records in the software you already run. It stages entries for a person on your team to review. That person clicks approve. The record posts. Nothing that touches money moves without a human in the loop.

That is what we mean when we say AI systems integrator. Not chatbots. Not a Zapier stack. A system built into the ERP, running against real records, with a person accountable for every write.

Ready to sanity-check your suspicion?

If you finished this list and recognised your operation in three or more of the signs, the next step is a conversation, not a purchase. Book a call with Throughline, or start with a fixed-fee AI Capability Audit ($7,500). The call is free. Call 865-417-3554.

About the author

Throughline is a small team of builders inside Keller Group. We build AI systems into our own operating companies first, then into yours.