Workflow automation consulting

Business Process Automation Consulting That Starts on the Floor, Not the Slide

Business process automation consulting is only worth paying for when it starts with the process you actually run, not the one drawn on a slide. Manufacturing Solutions Group is a small, operator-led firm in Texas. Our founder came up as a millwright on turbine decks and refinery floors before he started building and running companies, and he still builds the systems behind every engagement himself.

This page covers who it is for and not for, how an engagement runs week by week, what you own at the end, what moves the price, and how we check whether it paid off.

Is your business ready to automate, and when should you wait?

Most automation projects fail before they start, because the business was not ready and nobody said so. We would rather tell you to wait.

You are ready when three things are true. The process is repeatable, even if it lives in email and spreadsheets. Someone inside the company will own the result: one person who can say yes to a change and will notice when something breaks. And the pain is measurable in hours, errors, or missed revenue, not just a feeling that things are messy.

You should wait when the process changes every week, when the admin load is still one person's part-time job, or when nobody on your side will own the system after we leave. Then the right move is usually a process fix with no software, and we will say so.

Do not hire us if

  • You want a chatbot on the website and nothing behind it.
  • The process has never been done the same way twice.
  • Nobody on your team can give a few hours a week during the build.
  • You want a vendor to pick the tool first and the problem second.

What a workflow automation engagement looks like, week by week

A workflow automation consultant should be able to tell you what happens in week one and what you will be holding at the end. This is our run sheet. Timelines move with scope; the order never does.

The part most firms leave out is your time. Automation nobody on your side touched during the build is automation nobody on your side can run, so we fix a small amount of your team's time in every phase and put it in writing before we start.

Phase
1. Process walk
Typical length
Week 1
What we do
Watch the real process with the people who do it. Count handoffs, re-keys, waits.
What your team gives
3 hours from the people doing the work, 1 from the owner
What comes out
A map of how it really runs, with time and error cost per step
Phase
2. Fix, then design
Typical length
Weeks 2 to 3
What we do
Remove steps that should not exist. Decide what gets automated, simplified, or stays manual. Pick the lightest tool.
What your team gives
2 hours a week
What comes out
Future-state process, automation plan, system list, exception rules
Phase
3. Build
Typical length
Weeks 3 to 6
What we do
Build in your accounts, on your systems, with your data.
What your team gives
1 to 2 hours a week for access and quick answers
What comes out
Working automations in test, runbook started
Phase
4. Pilot on real work
Typical length
Weeks 6 to 8
What we do
Run the new process on live jobs beside the old one. Catch and fix the exceptions.
What your team gives
Daily use by the people doing the work
What comes out
Measured before and after on real volume
Phase
5. Handover
Typical length
Weeks 8 to 10
What we do
Train the system owner, finish the runbook, hand over every login.
What your team gives
A half day
What comes out
Everything in the next section

What you get at the end, and who owns it

You own all of it. The automations are built in your accounts under your logins. The documentation is yours. No retainer is required to keep anything running, and nothing is hosted on our side. If you want us around after handover, that is a separate choice, not a dependency.

Every engagement hands over

  • The before and after process map with measured time and error cost per step.
  • The automations, running in your systems, with the exception rules written down.
  • A plain-language runbook: what each automation does, what to check when it fails, how to pause it.
  • An access list of every account, login, and API key the system uses, held by you.
  • A support window after handover, and a 90-day review date set before we leave.

What drives the price

We do not publish a flat price for business process automation consulting, because the same request can be a two-week fix or a three-month build and a number on a page would be wrong for one of them. What we can do is tell you exactly what moves the price.

We work two ways. A fixed-scope project: we walk the process, write the scope, and quote a fixed fee for the build and handover, so the price is known before work starts. Or a phased build: the walk and design are a small fixed fee on their own, and the build is quoted from the design. We do not sell automation on a monthly subscription.

What moves the price up or down

  • How many systems the process touches. One CRM and one accounting system is a different job than five tools and a spreadsheet nobody admits to.
  • Data quality. Inconsistent customer records, part numbers, or job codes have to be cleaned first, and that is real work.
  • Approvals and exceptions. Every place a human has to decide adds a branch that has to be built and tested.
  • Whether the process needs fixing before it can be automated. Often it does. It is the cheapest part to skip and the most expensive part to have skipped.
  • Closed systems. Software with an open API connects quickly. Software that only exports a CSV, or has no export at all, costs more to work around.

Where automation fails: the five mistakes we get called in to fix

A good share of our automation work starts with something someone else built. The tool was fine; the thinking was not.

Automating a broken process: a step that existed only because of someone who left three years ago now runs faster, still wrong. Shadow spreadsheets: the official system is automated but the real work happens in a spreadsheet beside it. Brittle no-code chains: twelve zaps built by three people over two years, no documentation, and one changed field takes them all down. No owner after launch: the vendor left, the champion got promoted, the first failure went unnoticed for a month. Dirty data: the workflow runs perfectly on records that are 30 percent wrong, so it produces wrong output faster.

Automation on a shop floor, in a field-service shop, and in the back office

Most automation firms sell to offices. We spent years on the floor, so we start with the processes that run the physical business.

On a shop floor, the processes worth automating are the handoffs: quote to job ticket, job ticket to work order, work order to inventory and machine time, finished part to invoice. Every handoff is a place where someone re-keys information and something gets lost. Automating the handoff, not the machining, is where the hours are.

In a field-service shop, it is the loop around the truck: call, dispatch, quote, work, invoice, review request. Operators lose money not on the work but on the gap between finishing it and getting paid, and that gap is a process problem first.

In a professional-services firm or any back office, it is intake, document handling, and follow-up: the new-client form typed into three systems, the contract waiting in an inbox, the invoice reminder that never goes out. These happen every day, so an automation pays for itself in weeks.

AI agents, scripts, or RPA: which one fits which job

A business automation consultant in 2026 gets asked one question more than any other: should this be an AI agent? Usually the honest answer is no, and the firms selling agents will not tell you that. Our rule set:

Same job every time, clean inputs: a script or a simple integration. No judgment, so no AI. This is the cheapest, most reliable automation and most of what a business needs.

Clicking through old software with no API: classic RPA, robotic process automation. It works, it is fragile, and it should be a bridge to a better system, not a destination.

Judgment on messy inputs, like reading an email and deciding what it is about, pulling fields from a PDF that looks different every time, or drafting a reply for a human to approve: that is where an AI agent earns its place. We build agents for exactly that work, with a human check on every output that touches money, a customer, or a compliance record.

An agent is the wrong answer when the process is deterministic, the volume is too low to justify it, or nobody on your side can tell when it starts getting things wrong. An agent nobody is checking is a liability with a nice interface.

Plugging into what you already run

We build on the systems you already have. Replacing your CRM or accounting software is almost never the first move, and anyone who leads with it is selling software, not solving your problem.

Systems with an open API, including QuickBooks Online, HubSpot, Salesforce, Odoo, NetSuite, ServiceTitan, Microsoft 365, and Google Workspace, connect cleanly and support two-way automation. File-export-only systems can be automated on a schedule. Legacy software with no export can be automated at the screen level while you decide whether to migrate.

Build, buy, or consult: for one simple link between two modern tools, buy a connector or use a no-code platform yourself and save the money. For a process that spans several systems, has exceptions, or touches customers and money, the hard part is the design, not the tool, and that is when a consultant pays off. We have sent people away to do it themselves.

How we measure whether it paid off

Hours saved is the number everyone quotes and the easiest one to fake. We measure four things and baseline them before we build anything.

Time per cycle, from trigger to done, on real jobs. Labor hours per week on the process at the loaded rate. Error cost: how often the process produces something wrong and what each one costs to fix. Throughput: how many units the same team can handle per week.

We baseline in phase one, measure again during the pilot on real volume, and come back at 90 days. The 90-day number is the one that counts, because the novelty is gone and the exceptions have been found.

Why an operator, not a software shop, runs your project

Karl Gillihan, our founder and President, spent years as a millwright and projects lead on the turbine decks and refinery floors of the energy industry, at ExxonMobil and its XTO Energy operations, then Siemens Energy. He ran plant overhauls as a Project Manager at Universal Plant Services, then founded Stalwart Manufacturing Group, a precision parts shop serving oil and gas and energy customers, and ran it as CEO.

Today he runs MFGBase, a global procurement marketplace, and LocalAISource, a 50-state directory of AI professionals, and is a partner in ServiceStorm, a field-service platform. He builds the technology behind every MSG engagement himself. The person designing your automation has stood in a plant at 2 a.m. waiting on a part a broken process lost, and has written the software that stops it happening again.

Karl Gillihan

Karl Gillihan

President

Founder and President of MSG. He came up as a rotating equipment specialist in the energy industry and now runs MFGBase, LocalAISource, and ServiceStorm.

About Karl Gillihan

Frequently asked

How much does business process automation consulting cost?

It depends on how many systems the process touches, how clean the data is, how many approvals and exceptions there are, and whether the process has to be fixed first. We quote a fixed fee after the process walk, so you know the number before the build.

How long does a business process automation project take?

A single process across two or three systems typically runs six to ten weeks from process walk to handover, including a pilot on real work. Five or more systems, or a data cleanup first, takes longer, and we say so in the design phase.

What processes should a small business automate first?

The one that happens every day, involves re-keying the same information into more than one place, and costs you money when it is late or wrong. For most companies that is quote to invoice, new-customer intake, or scheduling and dispatch.

Do I need a consultant, or can I just use Zapier or Make myself?

One simple link between two modern tools: do it yourself. A process that spans several systems, has exceptions, or touches customers and money: the hard part is the design, not the tool, and that is where a consultant pays off.

What is the difference between business process automation and RPA?

Business process automation is the whole discipline: fixing and then automating how work flows through a company, with whatever tool fits. RPA, robotic process automation, is one technique inside it: software that clicks through screens the way a person would, mostly for old systems with no API.

Can AI agents replace our manual workflows, or is that hype?

Both. Agents fit work that needs judgment on messy inputs, like reading emails or pulling data out of inconsistent documents. They are the wrong tool for deterministic work, which is most of what a business needs automated. Where they fit, we add a human check on anything touching money, customers, or compliance.

Will automation work with QuickBooks, our CRM, or our ERP?

Almost always. Systems with an open API, including QuickBooks Online, HubSpot, Salesforce, Odoo, NetSuite, and ServiceTitan, connect cleanly. File-export-only systems can be automated on a schedule. Legacy software with no export can be handled at the screen level.

What do I get at the end, and who owns it?

You own everything. The automations run in your accounts, the runbook and access list are yours, and nothing is hosted on our side. No retainer is required to keep it running.

How do you calculate ROI on process automation?

We baseline four numbers before building: time per cycle, labor hours per week at the loaded rate, error cost, and throughput. We measure again during the pilot and at 90 days. The 90-day number is the one that counts.

Do you work with manufacturers and field-service companies?

Yes, that is where we started. Our founder spent years as a millwright and ran a precision parts shop, and he is a partner in a field-service platform. Shop-floor handoffs, dispatch to invoice, and back-office intake are the processes we automate most.

Want to know if your process is ready to automate?

Tell us what the process is and where it breaks. We will tell you whether it is a fix, a build, or a do-it-yourself job.

How do you want to reach us?