Home/Blog/Marketing Automation/How to Decide What to Automate First: A …
Marketing Automation

How to Decide What to Automate First: A Safe Order

How to decide what to automate first: pick a dropped handoff, check it works by hand, and start with internal actions before any message a customer sees.

Cover of the FIKR Space guide: How to Decide What to Automate First: A Safe Order

To decide what to automate first, choose a handoff that really gets dropped, confirm that you already do the process well by hand, and begin with internal actions before any message a customer will see. Selection matters more than construction.

The most consequential decision in automation is not how to build a workflow. It is which ones to build, and in what order. A well-built workflow aimed at the wrong thing is still the wrong thing, and it runs on a schedule. This post gives you a small set of filters you can apply to any candidate, in the order that keeps the cost of a mistake low.

How to decide what to automate first: four filters

Apply these to every candidate on your list, one after another. A candidate that fails a filter does not go forward yet.

  1. Does it protect a handoff that really drops? The candidate should come from a list you made by watching, not from a wish. If you have not made that list, start with how to find dropped handoffs.
  2. Does it already work by hand? If you have not done the process enough times to know what a good outcome looks like, you are automating a guess.
  3. Is it internal or outward? Internal actions come first, because they can be wrong without anyone outside the company knowing.
  4. Is it allowed? It must not be on your written list of things you will never automate.

Only a candidate that passes all four should be built. The list of candidates that fail is useful too: it tells you what to do by hand for now.

Filter one: the handoff has to be real

It is tempting to begin with the messages you wish you sent. A friendly sequence for new contacts is easier to imagine than a missing task on a deal. But the second is where business leaks. The guide is direct about this: start with the handoffs that get dropped, not with the messages you wish you sent.

These are the places where automation pays for itself at once, because the alternative is not a slightly worse version of the step. It is nothing at all. Write them down, rank them by how often they happen and how much each costs, and start at the top.

Free guide, 29 pages: The Automation Playbook. Send me the guide.

Filter two: the process must already work by hand

This is the rule that prevents most automated embarrassment. Do the process by hand first. Notice the exceptions, the cases where you deviated from the script and why. Notice which messages people replied to and which they ignored.

When you can describe the process in a few sentences and you no longer make judgement calls inside it, it is ready. Until then it is still a job for a person. The test is blunt, and it is worth repeating: if you cannot describe the process in a few sentences and you still make judgement calls inside it, it is not ready to be automated.

What you can sayWhat it meansNext move
"I describe it in two or three sentences"The process is stableMove to the next filter
"It depends, I decide each time"Judgement lives inside itKeep doing it by hand
"I have only done it a couple of times"You do not yet know what good looks likeDo it by hand more often
"The exceptions keep surprising me"The process is not yet understoodNote each exception and revisit

Filter three: internal before outward

An automation that creates a task for you, assigns an owner or notifies a teammate can be wrong without anyone outside the company ever knowing. An automation that sends a message to a customer can be wrong in public. The first is cheap to get wrong and cheap to fix. The second costs trust.

So the safe order is to begin with workflows that only touch your own team, in the form "when this happens, make sure the right person is told and has a task", and graduate to outward messages once you trust your triggers and conditions. Many of the most valuable workflows never send a customer anything at all.

Kind of actionWho can see a mistakeCost of being wrong
Create a taskYour teamLow, and easy to fix
Assign an ownerYour teamLow, and easy to fix
Tag a recordYour teamLow, but later steps may depend on it
Notify a teammateYour teamLow
Send a message to a customerThe customer, in publicTrust

Tagging deserves a note. It looks harmless, but a wrong tag can start other things later if another rule reads it. Treat the trigger that reads a tag as seriously as the one that applies it.

Filter four: the list of things you will not hand over

Write this list before you begin, because the temptation grows as the tooling gets easier. The usual entries are complaints, cancellations, bad news, anything posing as a personal note from a person who did not write it, negotiation, and the first real conversation with a serious prospect. The reasoning and the edge cases are in what should you never automate.

Decide the exit and the owner at the same time

Choosing what to automate also means choosing when it stops and who is responsible. These two decisions are easy to postpone and expensive to skip.

  • The exit. Every workflow needs an answer to the question of when a person should stop receiving it. A chase should stop when they reply. A nudge to book a call should stop when they book. An onboarding series should stop when onboarding is done. Founders design the start carefully and forget the end, which is how people end up receiving "just checking in" after they have already bought.
  • The owner. A workflow without an owner is a process nobody is responsible for, running on its own, indefinitely. In a very small company the owner is often you, and that is fine as long as it is a decision rather than a default.

If you cannot name the exit or the owner, the candidate is not ready. Park it and take the next one.

FIKR Hub: see FIKR Hub.

Measure before you switch it on

One more decision belongs before the build: what outcome this workflow should improve, and what that outcome looks like today. If it is meant to speed up first replies, record how long first replies take now. If it is meant to stop deals stalling without a next step, record how many stall now. Without a starting figure, you will never know whether the rule changed anything, and you may build three more before you find out it did not.

A worked example built from the guide's own words

Suppose your week of noticing shows that a new enquiry waits too long for a first reply. Test it against the filters.

FilterAnswer
Real dropped handoff?Yes, it is on the list of common gaps
Works by hand?Only if you reply well when you do reply
Internal or outward?Both: an owner and a task are internal, a confirmation is outward
Allowed?Yes, provided the confirmation does not pretend to be a personal note

The sensible first version is the internal half: when the enquiry form is submitted, assign the person to you and create a task to reply within the working day. The outward confirmation can follow once you trust the trigger. How to choose that trigger is covered in how to choose a workflow trigger.

What a good first choice feels like

A good first workflow is slightly boring. It protects a gap you have already felt, it does something small and visible only to your team, and you could explain it to a colleague in a breath. If the idea excites you because of how clever it is, be wary. Cleverness belongs in the judgement you keep for yourself, while the machine is there to be dependable.

It also helps to remember what comes after the first choice. You will read what the workflow did, compare it with what you expected and decide whether the next one is worth building. Selection is not a single decision at the start but a habit you return to, and each workflow you add should be one you can still review.

Free guide, 29 pages: The Automation Playbook. Send me the guide.

Mistakes in choosing

  • Starting with outward messages. Earn trust in your triggers before you point them at people outside.
  • Automating a guess. If the process has never worked by hand, customers will find its flaws for you.
  • Building the interesting one. Interest in building something is not evidence that it is needed.
  • Forgetting the end. A workflow with a start and no goal will keep chasing someone who already said yes.
FIKR Hub: see FIKR Hub.

Frequently asked questions

How many workflows should I start with?

One. Run it, read what it did and measure whether it changed the outcome. The number you run should be one you can actually review, and a single workflow keeps that promise easily.

Can I start with a welcome message to new customers?

You can, but it is an outward message, so it belongs after the internal ones. A promised welcome that depends on the founder remembering is a real dropped handoff, and a task that reminds the right person may be a safer first version.

What if everything on my list is outward?

Look for the internal half of each. Most outward steps have an owner to assign or a task to create beside them, and those can go first.

How do I know I am ready to move to outward messages?

When the triggers and conditions behind your internal workflows have behaved for a while and you have read what they actually did. Trust is earned by evidence, not by the passage of time.

The full method, with the checklists, is in The Automation Playbook, a free 29-page playbook.

LG

Luis Goncalves

// Founder & CEO at FIKR Space

Three-time founder. Built and exited Evolution4All before this. Now building FIKR Space — the operating infrastructure underneath every innovation ecosystem (startups, accelerators, governments, investors). Lisbon-based, works global.

Claim your founding member spot.

Every founding member is hand-reviewed. Tell us who you are and how you want to co-create — we get back inside 48 hours.

Role
Job Title
Country
Cookies

We use cookies.

Essential ones keep the site working. The rest only run if you say so. No dark patterns, no pre-ticked boxes.Privacy policy

Co-create with us. It's about building it together.Co-Create With Us