Automation does not fix a process. It multiplies it. So the first job is always to find out what you are about to multiply, and whether your foundations can carry it.

Automation multiplies whatever it lands on

There is a line widely attributed to Bill Gates that every team buying automation should keep in view: automation applied to an efficient operation magnifies the efficiency, and automation applied to an inefficient operation magnifies the inefficiency.

The pull towards automating is understandable. Salesforce's State of Sales research has reported reps spending only around 28% of their week actually selling, and automation promises to hand some of the rest back. But automation does not fail loudly on a broken process. It runs the broken process faster, at scale, with a straight face:

In each case the automation worked perfectly. It just worked on the wrong thing. A diagnostic is how you avoid paying twice: once to automate the process, and again to unwind it.

Find your binding constraint first

In The Goal (1984), Eliyahu Goldratt introduced the Theory of Constraints: every system has at least one constraint that limits its output, and improving anything else is an illusion of progress. Speed up a machine that feeds the bottleneck and you just build a bigger pile in front of the bottleneck.

Marketing and sales work the same way. If tracking, creative and targeting are all strong but leads wait a day and a half for a reply, the result is capped by that response time, however good everything else is. That is why we never average readiness scores. A strong overall score with one very weak area is not a strong score. It is the weak area with good intentions.

Goldratt's five focusing steps translate almost directly:

  1. Identify the constraint. Ask: if this one thing improved, would the overall outcome actually move?
  2. Exploit it. Get the most from it as it stands before spending money.
  3. Subordinate everything else to it. Stop optimising parts that are not the constraint.
  4. Elevate it. Now invest, in automation, people or process redesign, to lift it.
  5. Repeat. A new constraint will appear somewhere else. That is what progress looks like.

Practical ways to find yours: follow real leads rather than the org chart and see where work piles up; look for the step everyone quietly works around with a private spreadsheet; and ask what would break first if your budget doubled tomorrow.

Readiness: what your foundations can support

The way we frame it, automation readiness is a ladder rather than a switch. At the bottom, data sits in separate platforms and nobody is sure which numbers to trust. Automation here underperforms, because it has nothing reliable to act on. In the middle, foundations are partly in place and automation works inconsistently. Higher up, measurement is trusted and someone owns automation, so it delivers repeatable value. At the top, foundations are strong enough to support autonomous workflows.

We assess where an organisation sits across a handful of dimensions: whether the people running campaigns can read and act on data without queuing for an analyst, whether measurement is reliable, whether campaigns follow consistent structure and naming, whether a named person owns automation, and whether the tools match the team's skills. Each captures a different reason automation succeeds or fails, and the weakest one caps the rest.

One point is easy to miss: maturity changes the shape of risk, it does not remove it. Low on the ladder, the risk is that automation underperforms. High on the ladder, the risk is that it acts quickly and at scale in the wrong direction before anyone notices. Guardrails matter more as you climb, not less. The most common mistake is buying tools for the rung you want to be on rather than the one you are on.

How deep should the automation go?

"AI-powered" now describes almost every tool, so it tells you nothing. The better question is what the tool does by itself when nobody is watching. We think of execution depth as a spectrum with four honest answers:

Match depth to readiness. Lower on the ladder, recommendations and simple, reversible rules are the right tools. When evaluating a vendor, ask to see the actual log of what the system changed last week, and how you would undo the last ten changes. Vague answers tell you a lot. Our piece on guardrailed AI automation goes further.

Strategic risk flags: problems no automation can fix

Some campaigns fail not because of execution but because the strategy asked for something the business could not support. We look for a small set of recurring mismatches, each cheap to spot before launch and expensive to discover mid-flight. Two examples: innovation-heavy messaging aimed at buyers whose main fear is being blamed for a risky choice, and a high-volume lead campaign feeding a sales team of two or three people. Others involve high switching costs, a thin content library behind an enterprise motion, and a category where buyers see little difference between you and the alternatives. Automating a campaign with one of these flags just reaches the wrong outcome faster.

Audit, sprint, scale

Once you know the constraint and what your foundations can carry, keep the delivery shape small. Our approach is to audit read-only, so nothing changes while you look, and finish with a short plan rather than a strategy deck. Then sprint on one workflow at a time over a few weeks, with templates and team training. Then scale through regular iteration, adding the next workflow only when the last one is proven.

One workflow at a time keeps each change small enough to verify and reverse, gets a first result in weeks, and builds trust as it goes. Reversibility also tells you how much diagnosis a change needs. In his 2015 shareholder letter, Jeff Bezos separated one-way doors, decisions that are hard to reverse and deserve care, from two-way doors that can be walked back. A reminder or an internal notification is a two-way door: just build it. Anything that spends money, contacts customers or changes data you cannot easily restore is a one-way door. Audit first.

Try this this week

  1. Pick one process you are tempted to automate. Pull the real numbers: the median time between steps and how often records drop out.
  2. Trace ten real examples end to end. Write down where each one waited or broke.
  3. Ask: if we automated this exactly as it runs today, would the break get faster? If yes, fix it first.
  4. Ask: would we bet money on the data underneath it? If not, do not let a machine bet money on it either.
  5. Sort your automation ideas into one-way and two-way doors. Build the two-way ones now.

The honest limits

Where to go from here

Our free audits each score a set of dimensions and name the weakest one as your binding constraint. The automation readiness audit places you on the ladder in a few minutes, and the Sales Speed Audit checks whether response time is what is holding you back. If you are ready to build, see our sales pipeline automation work, and make sure you can measure the result in pipeline dollars.

The full scoring model, readiness criteria and pre-launch checklists sit in the methodology we share with clients. Get in touch if you would like access.

Questions

Why audit before automating marketing or sales?

Because automation amplifies whatever process it runs on. If tracking is broken or response times are slow, automation makes the problem bigger and faster. A read-only audit shows what to fix first and which automations are safe to build today.

What is a binding constraint?

It is the single weakest part of a system that caps the output of everything else, an idea from Eliyahu Goldratt's Theory of Constraints in The Goal (1984). Improving anything other than the constraint feels productive but does not move the overall result.

When can I skip the audit and just automate?

When the automation is small, cheap and easy to undo, such as a reminder, an internal alert or a record tag. If it spends money, contacts customers or changes data you cannot easily restore, audit first.

Sources

  1. Theory of Constraints Institute: Eliyahu Goldratt's Theory of Constraints
  2. Theory of Constraints Institute: The Five Focusing Steps
  3. Amazon: 2015 Letter to Shareholders (Jeff Bezos)
  4. Salesforce: State of Sales

More from the method