What a Company Should Automate First, and in What Order

Most companies that ask what to automate first already own the software to do it. The licences are already paid for and half of them are connected; what is missing is a decision about which piece of work goes first. The question is not which tool to buy. It is which piece of work, moved from a person to a rule, would remove the most waiting without creating new errors. That is a question about the work, not the software, and it has a fairly stable answer.

The answer is stable because automation has one property that does not change: it does exactly what the rule says, to every record, every time. That is the whole benefit and the whole risk. If the rule is clear and the records are sound, the work stops depending on who is free. If the rule is vague or the records are wrong, the workflow applies the vagueness at scale. So the order in which you automate is really the order in which you make each piece of work safe to hand over.

A useful way to hold this is three layers. The foundations are the things every workflow will read from. The repeatable work is what workflows carry. Judgment is what they hand back. You work down the layers in that order, and you do not skip the first one because it looks unglamorous.

Three things have to be true before any routing, enrichment or reporting workflow is worth building. First, the CRM is maintained: one record per company, an owner on each, required fields agreed and filled, and a habit of keeping it that way. Second, there is a written description of who you sell to, plain enough that a new hire could sort a list of leads into fit and not fit with it. Third, every form on your website writes a lead to the CRM with the fields your sales team needs, and it does so every time.

None of these is automation. Each is the input automation depends on. A routing rule cannot assign an owner if the territory field is empty. An enrichment step cannot tell a good lead from a poor one without a customer description to compare against. A weekly report counts whatever the CRM holds, and if the CRM holds three copies of one company, the report is wrong in a way that looks precise.

If any of the three is missing, that is the first project, and it is a short one compared with what it prevents. Write down the record standard. Write down the customer. Fix the forms. A team that does this first usually finds that some of the work it planned to automate no longer needs automating, because the mess that created it is gone.

• One CRM record per company, with an owner, the required fields filled and a person accountable for keeping it so.
• A written description of your customer that a person could use to sort a list.
• Every website form writes straight to the CRM with the fields sales needs, verified monthly.

Once the foundations hold, the candidates for automation are easy to recognise. They are the tasks that happen every week, follow the same steps each time, and stop whenever the person who does them is on a call, on leave or simply busy with something more urgent. The renewal reminder that goes out only when the account manager remembers. The territory reassignment that is done by hand each time a rep joins. The list that is rebuilt from an export every Friday because nobody trusts the segment.

Four kinds of work pass this test in almost every company. Routing: a new lead is scored against the rules and assigned to an owner on arrival. Enrichment: a new record is completed from the sources you already use and matched against what is already in the CRM. Reporting: the numbers are collected the same way each week and written up without anyone rebuilding a spreadsheet. Reminders: follow-ups, renewals and stale-record checks are raised on a schedule rather than from memory.

Within those four, start with the one that stalls most often and costs the most when it does. For most companies that is routing, because a lead that waits is a customer talking to someone else. Reporting is usually second, because it consumes a predictable block of someone’s time every period. Enrichment and reminders follow, and each one is easier because the previous one cleaned up the records it reads.

Some work looks repeatable and is not. Whether to discount for a customer who is comparing quotes. Which of two sales reps should own an account that spans both territories. Whether a sudden drop in form fills is a tracking fault or a real change in demand. Each happens often, and each time the right answer depends on something the rule was not written to see. Automating them does not remove the judgment. It hides it inside a rule that nobody reviews, and the first sign of trouble is a customer who noticed.

The practical rule is that automation carries the work up to the point where a decision is needed, then hands it to a person with the context attached. A lead the routing rules cannot place goes to a queue, not to a guess. A record with conflicting data is flagged, not merged. A message that will reach a contact waits for approval. This is not a limitation to work around. It is the design that makes the rest safe to run.

It also changes what the people on your team spend their week on. The waiting, copying and rebuilding go to the workflow; the parts that needed them stay with them. Most teams find the exceptions queue is a better use of a Tuesday morning than the inbox was.

Take one piece of work you are considering automating and walk it through the questions below in order. Stop at the first one you cannot answer yes to; that is where the work is. If you reach the end, the task is ready, and you can build it with some confidence that it will still be running correctly next quarter.

Run the same questions across your whole list of candidates and a ranking appears on its own. Tasks that fail at the foundations go into one group, to be fixed together. Tasks that pass everything are ranked by how often they stall and what the stall costs. Tasks that fail on judgment come off the list and become a queue for a person. That is the entire method; the value is in applying it honestly rather than in adding steps.

• Do the records this task reads live in a CRM we actually maintain?
• Could a new hire do this task correctly from a written page, with no one to ask?
• Does the task have a clear trigger, a fixed output and a place the output goes?
• Does it happen at least weekly, and does it stall when its owner is busy?
• Would a wrong result be caught by a person before a customer saw it?
• Is there a named person who approves the rule and will hear when it changes?

It is hard to see your own foundations clearly, because the people closest to the CRM are the ones who have learned to work around it. Three signals put a company on the foundations layer: nobody can name the owner of every open lead without looking; last month’s report and the CRM disagree on how many leads arrived; and at least one rep keeps a private spreadsheet because the CRM is not trusted. If any of these is true, that is the first project. The Marketing Automation Assessment on this site asks ten questions that surface the same signals and gives a readout on the page.

If the foundations hold, pick the one workflow from step two that stalls most often and build only that, with a named approver and an exceptions queue, before adding a second. How Aryeh Algorithms runs that work as a service is described on the Marketing Automation solution page; this piece is the order, not the offer.

Whichever route you take, the order does not change. Foundations, then the weekly work that stalls, then a clear line around judgment. Followed in that order, the list of workflows stays short and each one is easy to explain. Skipping the first step means automating the mess.