The New Enterprise Skills for 2026: From Platform Customizer to Agent Orchestrator

This spring I watched six keynotes back to back — SAP Sapphire, Microsoft Build, Salesforce Connections, Oracle AI World Tour, ServiceNow Knowledge and Workday DevCon. Six corporations, six stages, six marketing departments. And yet in 2026 they all told the same story: the platforms on which companies run their processes are becoming acting agent systems.

What really stayed with me, though, is not the product list. It is the question no vendor answers openly: who is actually supposed to operate this? Because when the software changes this fundamentally, so does the profile of the people who make a living with it. In 2026 the scarce skill is no longer to customize a platform. It is to orchestrate, ground and govern agents on top of it.

I have spent years helping organisations select enterprise software, roll it out and extract value from it. From that perspective I want to lay out which skill shift these keynotes really announce — and what it means for your workforce planning.

Six keynotes, one signal

The common thread is quickly told. SAP turned BTP into the Business AI Platform and presented an Autonomous Suite with, by its own account, 224 agents and 51 assistants. Salesforce proclaimed the “Agentic Enterprise”, Oracle the principle “built in, not bolted on”, ServiceNow positioned itself as the “AI Control Tower for business reinvention”, Microsoft built a control plane for agents with Agent 365, and Workday opened its platform to developer agents. Six times the same sentence in six dialects: system of record becomes system of action.

I have described this strategic convergence elsewhere in more detail — using SAP Sapphire 2026 and the shift from selling software to selling outcomes in “From SaaS to GaaS”. Here it is about the consequence at the level of people. I see four new skill axes — and a foundation that grows more important, not less.

Skill axis 1: Agent engineering and orchestration

In 2026 every vendor rolled out a workbench for building agents: SAP Joule Studio 2.0, Microsoft Copilot Studio, Salesforce Agentforce, Oracle’s Private Agent Factory, Workday’s Developer Agent and Orchestrate. The message everywhere: “Build agents in natural language, low-code to pro-code.”

That tempts people into the fallacy that no one is needed for it anymore. The opposite is true. Whoever builds an agent that approves orders or closes tickets on its own makes decisions about intents, guardrails, escalation paths and, above all, evaluation — the question of how you measure whether the agent is good enough to be let loose on real business processes. SAP explicitly calls its approach “intent-based, eval-based”. This is not a clicking job, it is engineering. The new role is no longer “consultant for module X” but “agent engineer for platform X” — and it demands process understanding and a feel for the probabilistic behaviour of language models.

Skill axis 2: The data and semantic layer

The sentence that came up in almost every keynote was, in spirit: “The language model doesn’t know your business data.” Oracle reduced it to “It all starts with data. Your data.” SAP built a context layer with a knowledge graph and “Company Memory”, Salesforce a semantic layer over Data 360, Microsoft the IQ layer over web, business and work data. The shared pattern: without a clean, semantically described data foundation, every agent hallucinates — it just sounds more convincing while doing so.

This turns a discipline long treated as background work into something business-critical: modelling ontologies and semantics, defining shared metrics, making data available without heavy copying. Whoever masters this layer decides whether a company’s AI is trustworthy. Why context is the real API of the AI era I explored in “Context Is the New API”.

Skill axis 3: AI governance and trust engineering

The most pointed sentence of the entire season came from ServiceNow: governance is “not a feature, it’s the whole ball game”. The image behind it — an agent deleting production databases and backups in nine seconds — stuck. Accordingly, every vendor built a control plane: ServiceNow’s AI Control Tower with a kill switch, SAP’s AI Agent Hub, Microsoft’s Agent 365, Workday’s Agent Passport, Oracle’s move of security down to the data layer.

A wholly new job profile emerges here. An agent needs an identity, access rights, an audit trail and a compliance review — in essence you treat it like an employee. Whoever sets that up is neither a classic security specialist nor a classic platform consultant, but a third thing: an AI governance specialist. In my advisory practice this is the role companies most often overlook — and the one that ultimately determines the speed of the entire AI rollout. How autonomous execution and human control can be balanced cleanly I explore in “Claude Managed Agents”.

Skill axis 4: Integration over open protocols

Three acronyms appeared in all six keynotes: MCP, A2A and, in part, A2UI. Agents will increasingly cross vendor boundaries — Workday agents show up in Microsoft Copilot, Salesforce capabilities run “headless” outside Salesforce, ServiceNow’s Action Fabric opens the platform to any AI. A notable aside: Anthropic’s Claude is a partner at five of the six vendors — a strong indication of the direction in which the market is consolidating.

For the classic integration roles — MuleSoft, Workday Studio, ServiceNow Integration Hub — this means a protocol upgrade, not a replacement. Whoever integrates systems must now understand how agents call tools and inherit context via MCP. Why context becomes the API of the AI era I covered in “MCP Explained”.

The foundation remains — and grows more valuable

For all the euphoria about agents, the most dangerous mistake would be to treat platform and process knowledge as obsolete. SAP called the ERP “the brain” on stage — processes, millions of data fields, authorisation logic. Workday framed the tension between “Enterprise Truth” (payroll, ledger — zero error tolerance) and “agentic velocity”. The point: an agent is only ever as good as the process model it sits on.

In concrete terms: the seasoned SAP FI/CO consultant, the Salesforce administrator, the ServiceNow developer — they are not becoming redundant. They become the indispensable foundation on which the new axes are built. Pure agent knowledge without domain depth produces exactly the expensive, plausible-sounding errors the vendors themselves warn about.

The skill matrix: old role, expanded role

If I condense the shift for decision-makers into a table, it looks like this:

SAP module consultant → SAP Autonomous Suite consultant: process knowledge plus Joule agents, KPIs and change management.
Salesforce administrator → Agentforce consultant: CRM configuration plus agent design on Data 360.
ServiceNow developer → AI specialist builder: workflows plus autonomous, governed “digital colleagues”.
Data engineer → semantic layer engineer: pipelines plus ontologies that let AI reason.
Security specialist → AI governance specialist: identity and audit for humans and machine agents.
Integration developer → MCP/A2A integrator: interfaces plus agentic protocols.

None of these roles replaces the old one. Each adds a layer that did not exist twelve months ago.

What this means for IT decision-makers

I draw three conclusions from this season. First: the bottleneck is not technology but the workforce. ServiceNow cited a demographic gap of up to 50 million workers by 2030 on stage — agents are the answer to that, but someone has to build and supervise them. Second: the most honest figure of the season also came from ServiceNow — 95 percent of companies cannot quantify the ROI of their AI. That is not a technology problem but a competence and governance one.

Third, and this is the real decision for you: build or buy? The four new skill axes cannot be grown internally overnight, and the market for them is swept clean. Whoever waits until their own team is fully reskilled gives away the window. That the real hurdle is almost never the technology but the organisation — that is where my work begins.

The vendors delivered their agent platforms in 2026. The open flank is the human side: which roles do I build, which do I buy, and how do I keep agents from becoming an ungoverned shadow system? If you want to think these questions through concretely for your organisation, get in touch — that intersection of strategy, platform and enablement is exactly where I work with companies.