Where to Start with Automation? A Prioritisation Framework for Businesses

Every business has plenty of ideas for automation: the phone, appointment scheduling, payment reminders, quotes, documents. The real question is where to start with automation. This article presents a simple prioritisation framework with three dimensions and a risk filter, explains the scoring and works through it for a law firm with six candidates.

Why the order matters so much

The first automation project shapes how the team thinks about every one that follows. If it goes well, people become more willing to take on the next. If it fails because it was too big, too risky or simply unimportant, you will still be hearing “We tried that once already” years later.

There is a practical reason too. A small business usually has one person who looks after automation alongside day-to-day work. That person can properly handle one project, not five. Choose the wrong order and you tie up your scarcest resource on the wrong topic. The practical roadmap for digital transformation describes how projects like these fit into a bigger plan.

The prioritisation framework: three dimensions

Rate each idea on three scales from 1 to 5. At least two people should rate independently, for example the owner and the office manager, and then discuss the results.

Benefit

• 5 = brings in revenue directly or prevents lost orders
• 4 = regularly saves many working hours
• 3 = noticeably improves the customer experience
• 2 = makes internal processes somewhat more pleasant
• 1 = would be nice, but changes little

Try to put a rough figure on the benefit: how many hours a month, how many missed enquiries? That protects you from going on gut feeling.

Feasibility

• 5 = can be done with a ready-made solution and no programming
• 4 = ready-made solution plus a simple connection to the calendar or CRM
• 3 = needs some adjustments or clean data that does not exist yet
• 2 = complex requirements, several systems, unclear processes
• 1 = not realistic at the moment

Urgency

• 5 = causes trouble or losses every day
• 4 = a known problem that comes up regularly
• 3 = gets in the way now and then
• 2 = will only become a problem as you grow
• 1 = purely precautionary

The risk filter before scoring

Before you calculate anything, ask one control question for each idea: what happens if the automation makes a mistake? A wrongly booked appointment is annoying but quickly fixed. Wrong information about a deadline, a medical finding or a legal question can do real damage. Ideas with a high potential for harm do not automatically end up in the bin, but they are not candidates for the first project and they need human oversight.

Check data protection at the same time. Anyone who handles professional secrets, such as medical practices, law firms or tax offices, must contractually bind service providers to confidentiality and sign a data processing agreement. For lawyers in Germany, professional law expressly governs the use of service providers (§ 43e BRAO, the Federal Lawyers' Act).

The scoring matrix: how to calculate

Multiply the three values: benefit times feasibility times urgency. The maximum is 125. Multiplying makes sure that one very weak dimension cannot be offset by two strong ones. An idea with a feasibility of 1 will therefore always stay near the bottom, however useful it would be.

Score	Recommendation
40 or more	tackle now, provided it passes the risk filter
20 to 39	schedule for next quarter
below 20	put on the ideas list, re-assess in six months

Example: a law firm decides where to start with automation

The ratings in this example are invented for illustration. A law firm with three lawyers and four staff collects six ideas.

Idea	Benefit	Feasibility	Urgency	Score
Answering calls and booking initial consultations	4	4	5	80
Structured intake of new client enquiries (area of law, opposing party, urgency)	4	3	4	48
Requesting missing documents from clients	3	4	3	36
Payment reminders for unpaid invoices	3	5	2	30
Website chat for general questions	2	5	2	20
AI drafts of legal briefs	4	2	1	8

Answering calls is clearly in the lead. In many law firms the phone rings while staff are typing up dictations or looking after clients at reception, and new clients simply call the next lawyer. It passes the risk filter as long as the AI handles only organisational matters and gives no legal assessment. An example scenario on client acquisition walks through how a law firm can win more client enquiries this way.

Structured client intake comes second, with one caveat: the conflict-of-interest check remains the firm's job; the AI only gathers the details for it. The brief drafts fail both the scoring and the risk filter. The article AI for law firms covers more ways AI can be used in law firms.

The website chat is revealing as well. It would be the easiest to implement, yet it lands right on the line for next quarter, because hardly anyone at the firm suffers from missing chat answers. This is where the matrix shows its real value: it stops a project from going first just because it is convenient. If two people's ratings for an idea are far apart, a short conversation about it is worthwhile. Usually one of them has seen something the other does not know about.

Typical mistakes when prioritising

• The most exciting project first. Technically appealing ideas often end up at the top even though they achieve little. The matrix forces a sober assessment.
• Overestimating feasibility. If the data you need is scattered across Excel lists and people's heads, an idea is not “easy to implement”.
• Ignoring existing processes. Automation should fit into the processes you already have instead of overturning everything. The article Integrating AI into existing processes has tips on this.
• Never re-assessing. What sits on the ideas list today may be urgent in six months, for example if an employee goes on parental leave. The skills shortage shifts priorities like these quickly; see AI as an answer to the skills shortage.

Frequently asked questions

How many ideas should I assess?

Five to ten is a good range. With fewer there is not enough to choose from; with more the group loses track. Collect the ideas with the team beforehand, because the people on the phone and at reception often know best where things get stuck.

What if two ideas have the same score?

Then feasibility decides. The project that is easier to implement delivers results sooner and builds trust for the next one.

Does the first project have to involve AI?

No. Sometimes the best starting point is a simple rule, a template or an online booking calendar. AI pays off where language has to be understood, for example on the phone or with freely worded enquiries.

When should I re-assess the list?

At the latest once the first project is running smoothly, and roughly every six months after that. Changes in the team, new customer channels or a switch of systems are also a good occasion.

Conclusion

The best way to decide where to start with automation is with three questions and one control question: how big is the benefit, how easy is it to implement, how urgent is it, and what happens if it goes wrong? Multiplying the three values quickly separates good starter projects from ideas for later.

For law firms where answering the phone comes out on top, the page on the digital legal assistant for lawyers describes one possible starting point.