Customer onboarding, the three Ps of a meeting, and coaching with a rubric: Harris Clarke of GuideCX

Harris Clarke's route into customer onboarding software runs through the State Department, which turns out to be less of a detour than it sounds. The through line is one thing: creating an experience deliberately.

Clarke is COO of GuideCX, the client onboarding platform, having been there since the beginning. Before that he was at Domo, before that at an insurance broker, and before that at the Department of Justice and the State Department.

What protocol teaches you about experience

At the State Department, Clarke worked in the Office of the Chief of Protocol, which handles how the United States engages with other countries. Not the what or the why or the policy, but the how. When there is a meeting, what does that meeting look like, where does it happen, what accompanies it.

His description of the job: creating the right experience, whether the meeting is tense or friendly, and being genuinely mindful about the experience you are creating.

The example he gives of what that means in practice is the checklist for a foreign leader visiting the president, which he describes as extensive. Including items an outsider would find absurd, such as whether a flag is angled a particular way, where the answer is that yes, it does need to be that way.

But the payoff is the part worth carrying into ordinary work. Having gone through it, you walk into the meeting with total confidence. You know exactly who is doing what and when, which means you can pivot if things do not go your way, precisely because you know what has been set up and what happens next.

The three Ps

Clarke's translation of that discipline into a technology company is the most immediately usable thing in the episode, and it is about meeting agendas.

If you send someone a calendar invitation, is it empty because you assume the title explains it? Particularly when you are involving people who were not expecting the meeting.

The framework he learned from someone years ago is three Ps.

Purpose. Why are we meeting.

Process. Who is doing what during the meeting.

Payoff. What we get out of having met.

His leniency about it is what makes it survivable. You do not have to literally write out the three headings, and you do not have to define everyone's role. But even a single-sentence agenda should embody why you are meeting, what will happen, and what you will have at the end.

His reasoning is about preparation rather than documentation. The last thing he wants is someone doing all their thinking inside the meeting. If he surprises someone with a question they could have had in advance, or if someone realizes mid-meeting what actually needs solving, the preparation did not happen.

And his argument for why it is worth the effort is cost. Meetings are expensive. Recurring meetings become expensive because you meet simply because it is on the calendar. Meetings with many people are expensive regardless of seniority.

So the three Ps do their real work on the organizer. Thinking through the purpose, the process and the payoff forces you to ask whether all these people need to be there, whether fifteen minutes would do rather than the default thirty, and whether the payoff is only relevant to a subset.

What bureaucracy is actually for

Michael's joke about government moving quickly sets up a genuinely useful answer.

Clarke's view is that bureaucracy has a deserved bad reputation and a real purpose. At the highest levels of government things need to move deliberately, because the implications of even slight policy changes are enormous.

What he took from it is not the process but the habit. Doing the work of figuring out what the systems should be, and then being ruthless about which ones you do not need.

His framing of the middle ground: make sure people understand the principles, then let them run. And be explicit about which systems you are sacrificing and why.

The Domino's pizza tracker for implementations

At Domo, Clarke worked across the post-sales side, and noticed the problem that became a company.

They implemented customers quickly. But with each new customer it could feel like starting over. Not literally forgetting what they had done, but with no templatized approach. If they had just built a particular use case for one large customer and another large customer wanted the same thing, there was no repeatable path, and a manual exchange of information about what happens next.

Then he met Peter Ord, who was starting a company around an idea Clarke describes as almost a Domino's pizza tracker for implementations. It clicked immediately, because Ord had no way of knowing that was exactly Clarke's frustration.

Michael, who is in Ann Arbor where Domino's is headquartered, adds the detail that makes the analogy better: Domino's operates as a serious data science company, using externalities like time of year and expected weather to determine the precise amount of dough each individual location needs.

And the point of the tracker, in Clarke's framing of why GuideCX exists, is emotional rather than logistical. Make sure the people involved are not anxious about what is happening, and know clearly what is happening.

Repeating yourself is underrated

Asked for a leadership lesson that changed how he works, Clarke names something most operators underrate.

Repeating your vision, and the strategy for reaching it, in multiple different ways.

His analogy is marketing, and it is a good one. Everyone accepts that a prospective customer needs many touchpoints before engaging, and companies spend heavily measuring that.

Leaders do not apply the same logic internally. You hold one meeting or send one email where you feel you have been clear, and that is simply not enough.

His prescription is to treat it like a campaign. You would not run only a post. You would have the article, the webinar, the website, the announcement, and outside noise. You saturate, because you do not want the point missed.

His closing line on it: if you have a clear vision and a good strategy, why fall flat on the one yard line and not do the extra work to make sure everyone is on the same page.

The part he does not think AI replaces

Asked what could never be replaced, Clarke acknowledges it may sound corny and names the human aspect.

Looking someone in the eye. Flying out to meet them. Understanding their needs beyond the metrics of the business.

His reasoning is grounded rather than sentimental. In business-to-business work, if it came down purely to a metric and a cost, there would be no need for human engagement at all. But you are working with people, which connects back to why GuideCX exists in the first place: making sure the people in a project are not anxious.

And he extends it internally. Whether someone feels valued, feels empowered to do their job, feels they are making a difference. There has to be some understanding beyond the work itself, of what is happening outside work.

Coaching with a rubric

Then, immediately, the thing he does think AI can take on, which is a genuinely good idea.

His advice is to start today. If you have in your mind what a perfect business review with a customer looks like, or the perfect support ticket, implementation call or sales demonstration, do the work of documenting it and building a rubric.

Ideally you already have it written. If not, write it so you could hand it to someone else and they could deliver the same coaching.

Then give the model the rubric alongside a call transcript or a support ticket, and ask it to grade against the rubric. Test the output, adjust where it does not grade the way you would, and iterate on the prompt.

His evidence is from that morning: he saw one come through for someone on the support team who was effectively coaching themselves with the tool, and the advice was exactly what he would have given. Including, he notes, the tone and delivery, which built the person up rather than discouraging them.

His framing of what you get: a second version of you, for anything where you can document how you would grade it. He mentions the persistent custom assistants both major providers offer for exactly this.

Michael's observation is the sharp one. Clarke named human interaction as irreplaceable, and coaching is a human interaction, yet this specific piece works.

Clarke's resolution is where the value actually lands. In the ideal case the AI coaching becomes the start of a conversation: do you agree with this, here is what stood out to me, here is what did not. If he does not spend his time reviewing every call and writing his own notes, he spends it reviewing the coaching that was given and discussing how to implement it.

Notebooks and footnotes

His other use is analysis. Weekly, monthly or quarterly updates uploaded into a notebook-style tool, then queried for trends.

The feature he singles out, specifically for people wary of AI, is citation. As you read a response you can click through each footnote to the source, which lets you judge quickly whether something is invented or accurate. His example: a claim that a metric trended a particular way with potential reasons, each footnoted, so you can check whether the source says what the summary claims.

What he values is the compression. Information that would previously have taken days or weeks to gather, available after a week of back and forth.

The annual strategy session

Clarke's answer to what the process should feel like is memorable: really hard, and then really inspiring.

The practical part first. You cannot do this at midnight on New Year's Eve and be ready on January 1st. You start at the end of the third quarter or the beginning of the fourth, take a hard look at the year so far, and let the rest play out.

And then the part that keeps it honest. Every company has challenges, from the largest technology companies down. So do not let your ego get bruised by the problems you find, because everyone has them.

What he learned the hard way

The lesson Clarke names is goal setting, which he says looks deceptively simple from the outside, whatever framework you use.

His warning: if you do not put significant time into figuring out what levers you are putting in place, you will accidentally put levers in place you did not mean to. In compensation plans, in one-on-ones, in the roles and job descriptions you hire for. And you get halfway through the year, realize you are badly off, and have only yourself to blame.

The work he describes doing instead is the full chain. These are the goals. These are the teams we have. These are the job descriptions those teams need, some of which may have to change. This is the management those teams need, the reinforcement in one-on-ones, the metrics to look at, the cadence for looking at them, and how we will know if we are on track.

And his conclusion about frameworks is worth hearing from someone who has tried several: there is no perfect framework for goal setting or strategic execution. What matters is sticking to it and being clear. If you are clear and internally aligned you can make real progress, and if you spend too long searching for the exactly right structure, you never will.

Being the steady hand

The story Clarke opens the episode with is the collapse of Silicon Valley Bank.

His point is that you do not plan for that and do not expect it. And that your role in the moment, fair or not, is to be the steady hand.

His honesty about what that costs is the part that makes it useful. Inside, you may be panicking, freaking out, wondering what happens next. What matters is that if people can look at you and see some level of confidence, they can have some confidence themselves.

The 5 things I took away from this conversation

1. Purpose, process, payoff. The cheapest meeting improvement available. Even one sentence covering all three forces the organizer to ask whether everyone needs to be there and whether thirty minutes is right. The work happens before the invite goes out.

2. Document the rubric, then hand it to the machine. Harris's coaching approach is the most concrete AI application I have heard from an operator. Write down what good looks like well enough that another person could coach from it, then let the model grade against it and spend your own time on the conversation that follows.

3. Say it as many times as you would say it to a customer. Nobody would run a campaign with one post. Leaders routinely announce a strategy once and assume it landed. The touchpoint logic applies internally and almost nobody applies it.

4. There is no perfect goal-setting framework. Harris has tried several and concluded the search itself is the trap. Pick one, be clear, stay aligned, stick to it. Time spent looking for the right structure is time not spent executing any structure.

5. Start the annual strategy in Q3. Not at the turn of the year. And go in expecting to find real problems, because every company has them, including the ones you admire.

FAQ

What is customer onboarding? The process of getting a new customer implemented and productive after they buy. Clarke's framing, drawn from the frustration that led to GuideCX, is that it should be templatized and visible rather than recreated for each customer, so nobody involved is anxious about what happens next.

Why does customer onboarding need to be tracked and visible? Because the experience is as much about anxiety as logistics. Clarke's comparison is a pizza tracker: the value is not only that the work happens, but that the customer can see it happening and knows what comes next without asking.

What should go on a meeting agenda? Clarke's three Ps: the purpose of the meeting, the process during it, and the payoff from having met. He accepts a single sentence covering all three, on the basis that the discipline mostly benefits the organizer, who is forced to justify the attendee list and the length.

How can AI be used for coaching? Document what excellent work looks like as a rubric, then have the model grade real artifacts, whether call transcripts or support tickets, against it. Clarke iterates on the prompt until the grading matches his own judgment, then spends his time discussing the coaching rather than producing it.

When should a company start its annual strategy process? End of the third quarter or beginning of the fourth. Clarke's point is that an honest look at the year requires time, and that you cannot compress it into the final days of December and be ready in January.

Also mentioned

• GuideCX, and the pizza tracker analogy behind its founding
• Domo, where Clarke saw implementations restart from scratch each time
• The Office of the Chief of Protocol, and what a state visit checklist teaches about experience design
• Domino's, headquartered in Ann Arbor, and its data science operation behind the tracker
• Persistent custom assistants in the major AI tools, used to hold a coaching rubric
• The collapse of Silicon Valley Bank, and being the steady hand while panicking inside

Listen to the full episode

Harris Clarke on Between Two COO's

Between Two COO's is hosted by Michael Koenig. Subscribe on Apple Podcasts, Spotify, or wherever you listen.