Your Approval Process Is Not Slow Because of People.

with Greg Poirier

In this video, Senior Vice President of the Salesforce Practice at Ateko, explains why enterprise leaders are asking the wrong questions about AI adoption. Instead of using technology to make legacy workflows faster, he details how organizations must audit their core processes to eliminate unnecessary approval gates and structural defects at the source. From replacing the “telephone game” of marketing briefs with direct AI agents to protecting human judgment where it matters most, this video breaks down what it takes to build an AI-native operating model that keeps pace with modern competitors.

Transcript:

Greg:

Hi, everyone. I’m Greg Poirier, Senior Vice President of the Salesforce Practice at Ateko

We’ve been thinking a lot about how artificial intelligence changes organizational structures, and here are some of the key observations that keep coming up in my conversations with clients. 
Many organizations I speak with are asking the wrong question about AI. 

They’re asking, how do we use AI to do the thing we do today faster?

It’s a reasonable question, but it’s also a shallow one.

The better question is, what processes were we running specifically to compensate for limitations that AI removes?

Because when you ask it that way, you realize a lot of what looks like organizational complexity,

is actually the plaque that builds up over time in big companies.

Most large organizations have approval architectures that would look completely foreign to a company being built today.

And when you trace that back to why each of these approval process gates exist,

you find the same thing every time.

It was solving a problem that AI has now made irrelevant.

So let’s take a marketing campaign brief as a concrete example.

A junior or intermediate team member is assigned to write the brief.

That’s the first problem.

The person writing the brief was never in the room when the business need was being identified,

and they are not the key stakeholder.

They’re working from secondhand knowledge and interpretation of what was said.

Every gate that follows is partially compensating for that translation lost.

Then the brief goes to their manager for edits and approval. Then to a requesting stakeholder.

Then maybe the senior marketing. If the spend is significant, the CMO gates on it as well.

Four rounds of approvals before a single piece of creative has been touched.

In some organizations, the brief itself then generates another round of creative feedback.

Each of these gates is doing one of two things.

They are either catching the translation error from the original step,

or they are performing a compliance check to make sure the campaign stays within brand, budget, and whatever guardrails the organization has established.

Both of these functions make complete sense given the tools available historically.

Both of them are now also solvable at the source with AI.


The approval structure was not built because people are bureaucratic by nature. It was built because the process had real structural defects, and approvals were the mechanism for management.

That distinction matters mostly for what comes next.


Well, the shift is upstream, not downstream. Instead of adding more gates at the end of the process, you fix the defects at the beginning.

The model that I find most compelling, an AI agent that interviews the stakeholder directly.

It asks the right questions, pushes back when the marketing brief is unclear and it’s been trained on the organization’s guardrails from the start. It can’t produce a marketing brief that goes off brand or off budget, because those constraints are baked into its intake logic.

The stakeholder reviews and tweaks the output, then the brief goes directly for execution.

What you’ve eliminated is the translation layer entirely.

The stakeholder who owns the business need is now also the person who generates the brief.

With AI doing the structuring and compliance work in real time.

The telephone game that was required for three approval gates to correct is gone at the source.

You don’t end up with zero approvals.

You end up with two, but they’re happening much further down the line at the point where human judgement is genuinely irreplaceable.

Does this final creative actually represent the brand at the level we want to put in front of our customers?

This is a judgement call.

That should stay human.

What should not stay human is the four rounds of approvals that preceded it for reasons that no longer exist.

So what does this mean for how senior leaders should be thinking about AI adoption right now? The reframe I keep pushing on. 
Stop auditing your workflows for where AI can be inserted, and make what you do today faster. Start auditing them for what they were built to compensate, and is no longer required.

Those aren’t the same audit.

I see this contrast most clearly when I work with early stage founders who are building AI native from day one.

They don’t have to unlearn legacy processes, and they don’t have an approval gone for it, like what we talked about previously.

They’re not designing faster horses. 
They’re not designing horses at all.

That is the competitive pressure that large organizations are not fully reckoning with yet.

The cost of the delay here is not abstract.

Every quarter you run that old approval architecture,

you’re paying for the overhead built to solve problems that are already solved.

That’s not just inefficiency.

That’s an opportunity cost that compounds, because your competitors, who are rebuilding from first principles, aren’t paying it.

So the question worth sitting with is,

if you were rebuilding your organization’s core processes from scratch today,

knowing what AI can do, would you design what you currently have?

Almost no one would.

That gap between what you would build and what you are running is the real AI transformation agenda.