While Everyone Talks Layoffs, Accenture Talks Promotions

Today, SAP Sapphire kicks off in Orlando.

Christian Klein is about to take the stage and announce SAP AI Foundation. A unified layer that connects SAP's own models with external models: OpenAI, Anthropic, Google Gemini, all of them. On SAP data. Governed by SAP.

The official narrative is that SAP is opening its ecosystem. The alternative narrative — the one some consultants are already publishing — is that SAP is installing new bars on the cage.

Both are true at the same time. But that's not the point. The point is what you're going to do when Klein's keynote ends. Because while the rest of the industry debates whether SAP is opening or closing, you have a different job. And that job needs a name.

The Operator

There's a scene in The Matrix that almost nobody mentions, but that defines the entire system. When Neo, Trinity, and the others enter the simulation, there's someone sitting in the hovercraft, in front of several monitors with green code cascading down. His name is Tank. And later, Link.

Tank doesn't enter the Matrix. He doesn't fight. He doesn't run. He doesn't dodge bullets.

Tank loads programs into Neo's mind. When Neo needs to fly a helicopter, Tank loads the piloting program. When Trinity needs karate, Tank loads the karate program. When someone needs to recognize a pattern in code scrolling at 200 km/h, Tank activates it from the outside.

Tank is not the hero. He's the Operator.

He's the one with full system visibility, the one who decides what capability goes in when, the one who orchestrates those in the field. Without the Operator, the subversive agents are blind. With the Operator, they're lethal. That's the most accurate metaphor I know for what it means to work in Technology in 2026.

The Difference from MacGyver

If you're over 35 and watched any TV, you probably remember MacGyver. That 80s show where the protagonist solved any problem with what he had in his pocket. A Swiss Army knife, duct tape, a paperclip, a lighter. He never used a weapon. He always used judgment.

The Operator's spirit is the same as MacGyver's. The problem solver. The one who looks at the room and says "with what I have here, I can build something."

But the Operator's practice is Tank's. Because MacGyver worked alone. With his two hands. In 2026, you don't just have two hands. You also have access to Claude, Gemini, Joule, multiple MCPs, dozens of specialized AI agents.

The modern senior consultant is no longer just a MacGyver. They're Tank with MacGyver's spirit. The one who knows which agent to load, when to load it, what data to feed it, which output to review, and when to discard everything and start over.

And this isn't theory. It's exactly what SAP is about to announce in Orlando.

What's Coming at Sapphire

The centerpiece of the announcement is called SAP AI Foundation.

It's a unified infrastructure that connects SAP's own AI models (Joule and derivatives) with external models: OpenAI, Anthropic, Google Gemini, Cohere. All of them. Operating on SAP data, in a layer that SAP governs.

Klein said it explicitly in his last earnings call: "infuse 50 years of domain knowledge into AI agents and govern the agentic layer for our clients."

Translated: SAP understood it can't compete against frontier models. So it decided to become the gateway where those models plug into SAP data. You don't call Anthropic directly. You call SAP's layer, which calls Anthropic, which returns to SAP's layer, which returns to you.

SAP isn't opening. SAP is becoming the toll booth.

This is the opposite of what Tobias Hofmann wrote last week on his blog. Hofmann argues that SAP doesn't innovate out of love for the customer — it innovates to stay relevant. To have a seat at the table where strategic decisions are made at Fortune 500 companies. And the SAP API Policy, that document we analyzed two weeks ago, isn't protection. It's controlled exclusion.

There are other SAP voices that went viral last week defending the API Policy. The argument — which I respect — is that SAP built the ecosystem over decades and has the right to defend it from predators who come to copy it, wrap it with AI, and sell it back. The phrase was: "I support AI. But I do NOT support predators."

Both positions can be true. SAP has the right to defend its ecosystem. And at the same time, SAP's decisions respond to SAP's interests, not always to its own customers'.

The point here isn't choosing sides. It's understanding the full play so you can make your own decision.

The Veteran Bonus, Revisited

Last week I talked about the Microsoft Research paper showing that AI corrupts 25% of work after 20 interactions. I talked about the veteran bonus — that market phase where the professional with accumulated judgment is worth more than ever because they're the only one who can detect the invisible error.

This week, that argument deepens.

The veteran bonus isn't just detecting errors. It's deciding the complete architecture. Which model to use for which task, which MCPs to connect, which agents to load, when to trust the output, when to discard it.

The veteran isn't the one who executes fastest. They're the Operator who decides what gets executed.

And SAP, with its AI Foundation, is building the perfect stage for Operators with judgment to shine and executors without judgment to be exposed.

When all your colleagues have access to Joule, Claude, and Gemini through the SAP layer, the difference won't be in who uses which tool. Everyone will have the same ones. The difference will be in who knows how to combine them with judgment.

What Accenture Just Admitted

As I write this, Accenture CEO Julie Sweet was giving an interview in Las Vegas where she acknowledged something almost nobody wants to say out loud. Her exact words: companies are investing massively in AI but struggling to translate that investment into measurable transformation.

The solution Accenture applied — what Sweet called "a playbook" — was tying internal promotions to each employee's AI fluency level. Not tool usage. Fluency.

That word matters. Fluency isn't knowing prompts. Fluency is knowing when to use what, with what judgment, on what data, for what purpose. Fluency is being an Operator.

Accenture, the world's largest consulting firm, just publicly acknowledged that the problem isn't AI. It's the judgment to use it. And it's reorganizing its entire operating model around that idea.

While the public narrative keeps talking about mass AI layoffs, Accenture is talking about promotions tied to AI fluency. Both things are happening in parallel, but they mean the opposite. One assumes the worker is replaceable. The other recognizes that accumulated judgment is promotable.

If Accenture is doing it, your company will be doing it in 6 to 18 months. And if you're not positioned as an Operator by then, you'll be the executor they evaluate — not the operator they promote.

What's Happening in the Trenches, Meanwhile

Microsoft publishes papers on AI's real capabilities. Accenture restructures its operating model. SAP announces AI Foundation. All of this happens on big stages, with lights and press coverage.

Meanwhile, in the professional forums where people talk without filters, the conversation is different. This week I read an architect on an enterprise project describe their week: they had to review two architecture documents and two test plans. All written with AI, with ten bullet points per section. And by Friday their eyes hurt from reading and reviewing every corner for details the AI missed. That the other developers also missed.

What bothered them most, they said, is that the work falls on the ones who have to review and approve these things. Everyone else gets comfortable with a few prompts, downloads pages and pages of generated text, and hands it to someone else to review. Forcing the reviewer to think twice as hard.

The post went viral because the complaint resonated.

On the other hand, veteran ERP analyst Jon Reed from Diginomica put it more bluntly this week. Reed expressed his frustration with the current narrative about AI agents replacing enterprise systems. And left a line that captures the entire problem in seven words:

Your probabilistic agent cannot complete deterministic workflows.

That line is the best technical description that exists today of AI's real limit in professional enterprise work.

A deterministic workflow is one where the result must be exact. Moving money between accounts. Closing an accounting period. Running payroll. Executing an SAP migration. Things where "almost correct" is the same as incorrect.

A probabilistic agent is what AI is. A machine that generates responses based on probabilities. Most of the time it's right. Sometimes it's not. And when it's not, it doesn't tell you.

Crossing those two categories without human judgment is exactly what the Microsoft Research paper showed fails. It's exactly what the architect I mentioned is suffering through when reviewing unfiltered generated documents. And it's exactly what Accenture is trying to avoid by tying promotions to fluency.

Two types of professionals are emerging right now. Those who use AI to generate volume fast without filtering the output. And those with the judgment to know where the probabilistic agent cannot complete the deterministic workflow.

The first group feels productive. The second group feels exploited. And the second group is who Accenture is promoting, who Reed is defending, and who your organization will need when the first group causes the first serious production incident.

If you're in the second group, you're not suffering the problem. You're the Operator.

The 3 Rules of the Operator

Rule 1: Know your models. Don't just use one.

Each model has different strengths. Claude is better for long-form reasoning and programming. Gemini is better for fact-checking. OpenAI models are better for structured tasks. Joule is better for native SAP context.

If you only know one, you're hostage to its weaknesses. The Operator knows the catalog and chooses by task.

Rule 2: Document your decision chain, not just your output.

When you solve something by combining three models and two MCPs, what needs to be documented isn't the final result. It's the path. Why you chose Claude for the first step, why you switched to Gemini to verify, why you discarded Joule's output.

That documentation is your proof of work as an Operator. It's what no CEO can automate yet. And it's exactly the type of fluency that Accenture just declared as the basis for promotion.

Rule 3: Build platform independence.

SAP AI Foundation is going to make it very easy to stay inside the SAP ecosystem. Technological laziness will push you toward using only the official connectors, the certified agents, the endorsed tools.

Keep at least one external MCP server active, one model outside SAP's gateway, one tool you control. Independence is expensive today, but the day SAP changes its API Policy again (and they will), that independence will save you.

The Questions for This Week

As Sapphire unfolds in Orlando, you'll see an avalanche of announcements. Joule reimagined. SAP AI Foundation. Partnerships with Microsoft, OpenAI, Anthropic. Agents for finance, supply chain, HR.

The question isn't what SAP announces. The question is what decisions you make as an Operator after hearing the announcements.

Will you stay inside the SAP ecosystem, depending on Klein's decisions?

Will you leave entirely, assuming you can operate without the SAP layer?

Or will you take the harder but more profitable position — the Operator who moves in and out at will, connecting the best of inside with the best of outside, with accumulated judgment and documented proof of work?

There's no correct answer. But there's your answer. And that answer is what will define your career over the next 24 months.

Those who understand the fundamentals orchestrate. Those who don't are orchestrated.

This week in Orlando, SAP will show you the stage where the game will be played. You decide whether you enter as a spectator, a piece, or an Operator.

Sources:

Julie Sweet, CEO Accenture, interview with Fortune at the Great Place to Work For All Summit, Las Vegas, May 10, 2026.

Jon Reed, veteran ERP analyst at Diginomica, statements published on LinkedIn, May 2026.

Public discussion in professional developer forums on Reddit, May 2026.

Tobias Hofmann, "It's about control," itsfullofstars.de, May 2026.

Christian Klein, SAP Q1 2026 earnings call.

Constellation Research, "SAP Sapphire 2026 themes," April 2026.

Sapphire Innovation Guide, sap.com.

Previous newsletters referenced: Mythos (the access gap), Shadow AI (the internal gap), SAP API Policy (the two SAP castes), DELEGATE-52 (the judgment gap).