Automation projects stall most often at the question of where to start. This is a concrete answer: fifteen processes that recur in nearly every business, follow a describable pattern, and have a clear definition of done — the three conditions that make a process a good first candidate.

Intake and first response

Every business loses something at the front door. These processes decide whether an enquiry becomes an opportunity or a missed call.

Intake is worth automating first in most service businesses because the cost of failure is immediate and measurable. An unanswered call does not wait politely in a queue; it goes to whoever answers next. Automating intake does not mean removing people from customer conversations — it means ensuring no conversation is missed while they are busy with the last one.

  • Answering inbound calls, including after hours and during overflow
  • Capturing and qualifying inbound leads from every channel into one routed queue
  • Booking, confirming, and rescheduling appointments against live availability
  • Routing enquiries to the right person with the context already summarised

Documents and data entry

Document work is the clearest case for automation because the input volume is high and the task is mechanical. It is also where staff time disappears least visibly.

The reason documents resist traditional automation is that their format varies constantly. Two suppliers send invoices carrying the same information in entirely different layouts. A rule written for one breaks on the other. Interpreting the document rather than pattern-matching its layout is what makes this category newly automatable.

  • Extracting fields from invoices, forms, and contracts into your systems
  • Classifying and filing incoming documents by type, client, or project
  • Comparing document versions and flagging what changed
  • Drafting recurring documents from templates and project data

Finance and administration

Finance processes tend to be deadline-driven and repetitive — a combination that suits automation and punishes manual handling.

They also tend to be seasonal. A team sized for the quiet months is overwhelmed at close or during busy season, and a team sized for the peak is underused the rest of the year. Automating the mechanical portion flattens that curve without hiring against the peak.

  • Requesting, tracking, and chasing missing client documents
  • Routing payables and receivables, including follow-up on outstanding items
  • Assembling recurring internal and client reporting from live data

Sales and pipeline

Pipelines leak wherever progress depends on someone remembering. Automating the memory is usually worth more than automating the selling.

CRM hygiene is the clearest example. Records go stale because updating them is nobody's priority in the moment, and a pipeline nobody trusts stops being used for decisions. Keeping records current automatically restores the data without asking salespeople to spend more time on admin.

  • Persistent, contextual follow-up on leads that have gone quiet
  • Keeping CRM records, activity, and stages current without manual entry

Operations and coordination

Coordination work rarely appears on anyone's job description, which is why it consumes so much time without ever being measured.

It is also the work that scales worst. Every additional project multiplies the status requests, approval chases, and vendor messages, and that load lands on the same few people. Automating coordination is often what lets a team take on more projects without adding administrative headcount.

  • Chasing approvals until they are resolved
  • Tracking project status and open items from the systems where work is recorded
  • Handling routine vendor correspondence and document exchange

How to choose the first one

Rank candidates by frequency multiplied by handling time, then filter for those with a clear definition of done. The highest-volume process is not always the right start; the right start is the one where success is unambiguous, because that is what earns support for the next one.

Resist the temptation to begin with the most painful process in the business. Painful usually means complex, and complex is a poor first project.

What to measure

Decide the measure before you deploy, and make it something you already track. Hours returned, response time, error rate, or throughput per person are all defensible. A baseline captured after go-live is not a baseline.

Frequently asked questions

Which process should we automate first?

The one that happens most often and has an unambiguous definition of done — not necessarily the most painful one. A clear first result builds the case for the next project, and complex processes make poor first attempts.

Can AI automate a process that is not written down?

Yes, but the first step is writing it down. Mapping how the work actually happens, including the workarounds people have invented, is usually valuable on its own and often reveals steps that can simply be removed rather than automated.

Will automating these processes eliminate jobs?

In most deployments the work that gets automated is work nobody was hired to do — rekeying data, chasing approvals, answering repeat questions. The common outcome is the same team absorbing more volume without adding administrative headcount.

How many processes should we automate at once?

Start with one, prove it against real work, then expand. Simultaneous rollouts across several departments make it hard to tell what worked and much harder to correct course.

What if our process changes frequently?

That is manageable, but it should be a design input from the start. Processes that change often need automation built with the variable parts configurable, rather than hard-coded into a fixed sequence.