How to Build an Automation Roadmap
Picture the Monday staff meeting. Someone says the invoices are late again. Someone else says the intake forms are a mess. A third person mentions that the weekly report takes half of Friday. Everyone agrees all three should be automated, and nobody agrees which comes first. By the end of the hour the loudest voice wins, a tool gets bought, and three months later the invoices are still late.
A roadmap is the thing that replaces that meeting. It is not a grand plan and it does not need a consultant. It is a ranked, dated list of automation work that you can defend to your owner, your finance lead, and your own team. If you are an operations director with a small staff and a limited budget, this is how I would build one.
Start with an inventory, not a wish list
A wish list records what people want. An inventory records what actually happens. The difference matters because people are poor at estimating their own repetitive work. They remember the annoying task and forget the three small ones around it.
For two weeks, collect every recurring task that touches more than one person or more than one system. Ask each team member to keep a simple log. Do not ask for a polished process document. Ask for a line per task with the following columns.
| Column | What to write |
|---|---|
| Task name | A plain verb phrase, such as "retype signed orders into the accounting system" |
| Trigger | What starts it, such as an email, a date, or a customer action |
| Frequency | Times per week or month |
| Minutes per run | Honest, rounded to the nearest five |
| People involved | Who touches it |
| Systems touched | Every app, spreadsheet, or inbox |
| Error history | Last time it went wrong and what it cost |
| Owner | One named person |
The owner column is the one people skip, and it is the one that stalls projects later. If nobody owns a task today, nobody will own its automated version either.
Expect between 30 and 80 rows for a company of 15 to 50 people. That is normal. You are not going to automate most of them. You are building a map so that the few you pick are picked on evidence.
Score each item on five questions
Scoring does not need to be statistical. It needs to be consistent, so that two different people scoring the same row land close together. I use five questions, each scored from 1 to 5.
| Question | A 1 means | A 5 means |
|---|---|---|
| Time saved | Under an hour a month | Over 20 hours a month |
| Error cost | Mistakes are cheap and caught fast | Mistakes reach customers or the ledger |
| Stability | The process changes monthly | The process has been the same for a year |
| Data readiness | Inputs are free text and scattered | Inputs are structured and in one place |
| Reversibility | Hard to undo if it misbehaves | Easy to switch off and return to manual |
Add the first two for value. Add the last three for feasibility. Plot each row on a simple grid with value on one axis and feasibility on the other.
Notice that stability and data readiness carry as much weight as time saved. A task that eats 20 hours a month but changes every few weeks is a poor first candidate, because you will rebuild it constantly. A task that eats six hours a month, runs the same way every time, and starts from clean data is often the better first win.
A worked example, with made up numbers
This example is hypothetical and the numbers are illustrative. Suppose a 25-person firm has four candidates.
| Candidate | Hours per month | Error cost | Stability | Data readiness | Reversibility |
|---|---|---|---|---|---|
| Retype orders from email | 22 | 4 | 4 | 2 | 5 |
| Weekly sales report | 16 | 2 | 5 | 4 | 5 |
| New client intake | 10 | 5 | 3 | 3 | 3 |
| Vendor bank change requests | 3 | 5 | 5 | 4 | 2 |
The order retyping has the biggest raw time number, but the data is messy because orders arrive as free text in email. The weekly report scores lower on value and much higher on feasibility. The vendor bank change process is small in hours but protects the company from payment fraud, so its error cost dominates.
A reasonable first quarter might be the weekly report, because it is a low risk way to build the habits of scheduling, monitoring, and review. The second item might be a structured intake form for orders, which is a data readiness project that makes the retyping automation possible later. The bank change verification can be a short, mostly procedural fix that needs very little software.
Notice what happened. The ranking is not the same as sorting by hours saved. The roadmap has a reason for each position, and you can say that reason out loud in a meeting.
Sequence by dependency before you sequence by score
Scores tell you what is attractive. Dependencies tell you what is possible. Before you finalize the order, draw a line from each item to whatever it needs first.
The usual dependencies are these.
- A clean identifier. Automations that match records need a key that is the same in both systems, such as a customer number or an invoice number. If two systems name the same customer differently, fix the identifier first.
- A system of record. If the same fact lives in three places, decide which one wins before any automation moves data between them.
- Access. Many automations need a service account or an API credential. Ask early. In some vendor plans the API is a paid tier, and that changes your budget.
- A named owner and a backup. One person who knows how it runs, and one who can step in.
- A place for failures to go. Every automation fails sometimes. Decide where the failed items land and who looks at them.
When an attractive item depends on an unattractive one, the unattractive one moves up. That is how foundations get funded without anyone having to call them foundations.
Put failure and fallback into every row
Each roadmap item should state what happens when it breaks. This is where small teams get hurt, because the person who built the automation is also the person doing a day job.
Write three short lines for each item.
- What does a failure look like, and who is notified?
- What is the manual fallback, and how long can the business run on it?
- How do we switch the automation off?
In Puerto Rico there is a fourth line worth adding. If power or connectivity is out for a day or more after a storm, which automations stop, and which of them must keep running? Cloud-hosted automations often keep working when your office does not, but only if the people who need to act on their output can reach them. A manual fallback that lives on a laptop in a closed office is not a fallback.
Fund it in small steps
A roadmap that needs one big approval tends to die waiting for it. Break the work into stages, each with its own small budget and its own stop condition.
A pattern that works well for small teams has four gates.
- Baseline. Measure the current time, error rate, and delay for the process. This costs staff time and nothing else. Without it you cannot prove the later savings.
- Pilot. Automate the narrowest useful slice, such as one customer type or one form. Run it beside the manual process.
- Expand. Widen the scope only if the pilot met the stated numbers.
- Operate. Move the item into monitored, owned, documented production with a named maintenance budget.
Every gate has a decision. Continue, change, or stop. Stopping is a legitimate result, and it is cheap when you stop at the pilot. Tell your finance lead this up front. It makes approvals easier, because the exposure at each gate is small and known.
Budget for maintenance from the first step. Integrations need credential renewals, vendor changes need rework, and staff turnover needs retraining. A common rule of thumb is to expect ongoing care to be a meaningful share of the build cost every year. I will not give you a percentage, because it varies with how many vendors are involved and how often they change their interfaces. Ask the person building it to estimate it in writing.
If you are comparing the effort against what you pay for software today, recurring vendor and license charges are separate from any services fee. Ozcorp InfraOps Base, for what it is worth, starts at $500 per month with vendor and license charges billed separately. That is only one reference point, and your own baseline matters more.
Decide what not to automate
A good roadmap has a "not now" list, and it is as valuable as the "do" list. Leave these off.
- Processes that are still being invented. Automating a moving target locks in last quarter's habits.
- Work that depends on judgment about people, such as a delicate customer call. Support it with reminders and templates, not a script.
- Tasks so rare that the build cost can never be recovered.
- Anything where the manual process is broken. Fix the process on paper first. Automating a bad process gives you a faster bad process.
Write the reason next to each item. Reasons age well. In six months you can reread them and see which have changed.
Keep the roadmap alive
A roadmap is a living document, and a quarterly review is enough for a small company. At each review ask four questions.
- Did the shipped items hit their baseline targets?
- Did anything new show up in the inventory that outranks the plan?
- Have any dependencies changed, such as a vendor replacing a system?
- Is any automation now unowned because someone left?
Retire anything that no longer earns its keep. Unused automations are quiet risks, because they keep running with credentials nobody remembers granting.
What to do Monday morning
You can start without buying anything.
- Send the inventory template to your team and give them two weeks.
- Name one person to merge the logs into a single sheet.
- Pick the first five rows by feasibility, not by volume, and score them using the five questions above.
- Write the failure and fallback lines for your top candidate.
- Book a 30 minute review with your finance lead to agree on the baseline and the first gate.
That is a roadmap in embryo. It will be imperfect, and that is fine. A rough ranked list that people trust beats a perfect one that never ships.
If you would like a second pair of eyes on your inventory or your scoring, we are happy to look at it with you. Send us the sheet and we will tell you honestly which two or three items we would do first, and which we would leave alone.
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.