Business Process Automation: Where a Small Company Starts
Finding the processes worth automating, costing the savings in hours and money, and the order to roll them out — without an expensive ERP.
A small business owner hears "process automation" and pictures a $30,000 ERP and six months of implementation. In practice the first three automations in a 5–20 person company cost $600–2,500 together and free up 15–30 hours a month. The difficulty is not technical — it is picking the right three, because automated chaos is still chaos, only faster. Below: how to find the candidates, a payback formula you can run yourself, and the rollout order.
15–30
hours a month freed by the first wave
$300
floor price for one genuinely useful automation
2–5
months to payback on a realistic project
3
processes maximum in the first round
Finding the processes worth touching
The worst method is gathering the team and asking what annoys them. People name what irritates them emotionally, not what eats hours. You need numbers, and collecting them takes two weeks and no tooling.
- 1
Two weeks of logging
Everyone who handles routine work logs it: what they did, how many minutes, how many times a day. Two weeks is enough for the patterns to surface.
- 2
Convert to monthly volume
Frequency × duration × headcount. A five-minute action performed 12 times a day by two people is 22 hours a month. A two-hour action performed monthly is two hours.
- 3
Drop anything that needs judgement
If it requires context, negotiation or ownership of a decision, it is not a candidate. Automation moves data extremely well and makes decisions extremely badly.
- 4
Check the process is needed at all
Half the daily reports we have seen were needed by nobody. Cancelling is always cheaper than automating: ask who opened that file last week.
| Process trait | Automate | Leave to a person |
|---|---|---|
| Frequency | Daily or several times a day | Monthly or rarer |
| Rules | Clear algorithm, no exceptions | Every case is unique |
| Input data | Already digital | Paper, voice, poor scans |
| Cost of an error | Visible and easy to roll back | An error costs you the client |
| Who performs it | Several people, each differently | One expert using their judgement |
Working out whether it pays back
The formula needs no consultant: (hours per month × internal cost per hour) − monthly tool fees = savings. Divide the build cost by that number and you get the months to payback. Anything over 8 months is a weak project — find a different process.
An example from our own work: two account managers each spent 40 minutes a day copying enquiries into a spreadsheet and typing the same replies. That is 26 hours a month, or $208 at an internal cost of $8 per hour. The build cost $700 plus $25 a month — payback in 3.8 months.
- Lost enquiries. The largest share of the saving usually is not in hours but in the 10–20% of enquiries nobody got round to.
- First response time. Cutting it from two hours to two minutes lifts conversion into a real conversation more than most tweaks to your landing page.
- Data entry errors. One wrong address or amount means a return, a goodwill credit and a bad review.
- Support cost. A minus, not a plus: budget 10–15% of the build cost per year, because integrations break whenever somebody else updates their API.
Where almost everyone starts
These five recur from a hair salon to a wholesale distributor: fast results, no reorganisation, and time for the team to get used to the idea.
| Process | What we do | Typical saving | Budget |
|---|---|---|---|
| Enquiry capture | Forms, email and messengers into one CRM with a source tag | 8–20 hrs/mo | $300–800 |
| Standard questions | A bot with an answer base and clean handoff to a person | 10–25 hrs/mo | $400–1,200 |
| Customer reminders | Automatic notices about appointments, payments, renewals | 4–8 hrs/mo | $200–500 |
| Internal alerts | A new enquiry or an overdue task into the team chat | 3–6 hrs/mo | $150–400 |
| Daily summaries | An automatic sales report in chat instead of a manual export | 4–10 hrs/mo | $250–600 |
The order is not arbitrary: consolidate the enquiry flow first, then build on top of it. Automating reminders while contacts live in three different places is wasted work. How to consolidate the data is covered in our guide to CRM integration.
Off-the-shelf, no-code or custom build
| Option | When it fits | Price | Main risk |
|---|---|---|---|
| Off-the-shelf SaaS | Standard job: mailing, bookings, till, subscriptions | $10–100 per month | Dependence on someone else’s pricing and roadmap |
| No-code (n8n, Make) | Wiring 2–4 services, up to a few thousand runs a month | $20–80/mo plus $200–600 setup | Cost scales non-linearly with volume |
| Custom build | Non-standard logic, high volume, sensitive data | from $800 one-off | You need someone to maintain it |
The line for moving to custom code sits at exception count: once a scenario grows a dozen conditions along the lines of "unless this is a returning customer", the no-code canvas becomes unreadable. Both approaches are compared in our piece on no-code versus a custom build.
A rollout order that does not break the business
- 1Describe the process as it is today on a single page: who does what, where the data comes from, where it goes. If you cannot write it down, it is too early to automate it.
- 2Record a baseline metric: average handling time, enquiry volume, percentage missed. Without a "before" number you cannot prove the effect.
- 3Launch on one team alongside the old way for two weeks — you find the discrepancies, not your customers.
- 4Write down what happens when the automation fails. Not "if" — when: third-party APIs go down, tokens expire, quotas run out.
- 5Roll out to everyone and remove the old way. While the old spreadsheet exists, half the team will keep using it.
Signs the automation has actually stuck
- Nobody keeps a parallel spreadsheet "just for themselves"
- One named person owns the scenario and knows where to look when it breaks
- Failure alerts land in the team chat, not in a log file nobody opens
- A new hire can be brought into the process in 20 minutes without dedicated training
Automation does not make a company more organised. It makes its existing level of organisation cheaper to execute — and far more visible when it is missing.
Frequently asked questions about process automation
Where should a 5–10 person company start with automation?
Start by consolidating enquiries into one place: every form, email and messenger should land in a single system with a source tag. That costs $300–800, takes one to two weeks, and immediately reveals how many enquiries were being lost. Everything else — reminders, bots, reports — is built on top of that data flow.
What does business process automation cost for a small company?
One straightforward automation, such as routing enquiries into a CRM with a chat notification, costs $300–800 as a one-off plus $10–50 a month in service fees. A package of three or four processes for a 5–20 person company usually fits within $1,500–3,000. Budget an extra 10–15% of the build cost per year for maintenance, because external services keep changing their APIs.
How do I tell whether an automation will pay for itself?
Multiply the hours the process consumes per month by your internal cost per hour, subtract the tool fees, then divide the implementation cost by the result. Under 6 months is a strong project, 6–8 is acceptable, over 8 means picking a different process. Add the value of lost enquiries, because that is usually where the biggest effect sits.
What should never be automated?
Processes that run less than once a week, require professional judgement, or work with paper and voice-note inputs. Also anything that cannot be expressed as rules without a dozen exceptions — maintaining that logic costs more than doing the work by hand. And separately: a process that could simply be cancelled should be cancelled.
In short
- Find candidates with numbers: two weeks of activity logging, then frequency × duration × headcount.
- Automate what is frequent, rule-based and already digital. Leave judgement, rarity and paper to people.
- Payback comes from one formula; anything over 8 months is a signal to pick a different process.
- Start by consolidating enquiries, then standard replies, reminders and reports — in that order.
Related reading
CRM Integration: How Leads Stop Falling Through the Cracks
Leads are not lost by people. They are lost by the absence of a system. Let us build one.
Telegram Bots for Business: 12 Scenarios That Save Real Hours
A bot earns its keep where a human repeats the same action fifty times a day — not where it looks modern.
Bot Builder or Custom Code: Where the Line Really Sits
A builder is cheaper right up until you need the one thing it cannot do.