What does 'agent-native' mean for a professional services firm?

Agent-native means your firm's operations are designed around AI agents that complete multi-step tasks autonomously — not just around AI tools that assist with individual tasks. The difference: a tool helps you write a client email; an agent gathers the client data, drafts the email, checks the history, and sends it based on a trigger. CBIZ's April 2026 announcement with Microsoft is the first public declaration by a top-10 professional services firm that this is their operating model.

What tools does the CBIZ agent-native platform use?

CBIZ's agent-native platform uses three Microsoft products: Microsoft 365 Copilot (the foundation layer for daily productivity tasks across 9,500+ employees), Copilot Studio (the automation layer for building firm-specific AI agents), and Microsoft Foundry (the orchestration layer for multi-step workflows that cross departments or systems). Each layer does a different job — Copilot handles day-to-day work, Copilot Studio builds the automation, and Foundry connects everything end-to-end.

Can a small professional services firm implement an agent-native approach?

Yes — and at dramatically lower cost than a firm like CBIZ. The same three-layer architecture (foundation tools, custom agents, workflow orchestration) scales to a 10-person firm. At that size, the foundation layer might be Microsoft 365 Copilot or Claude for Business. The automation layer is Copilot Studio, Zapier, or Make. The orchestration layer — connecting everything — is the step most small firms skip, but it's where the compounding return lives. You don't need 9,500 employees to start.

What is Microsoft Foundry and why does it matter for professional services firms?

Microsoft Foundry (part of Azure AI Foundry) is a platform for building, deploying, and monitoring enterprise AI agents — AI that executes multi-step workflows across systems rather than just answering questions. For professional services firms, the significance is that Foundry provides the infrastructure for AI that 'acts on your behalf': it can trigger a workflow, gather inputs from multiple systems, execute tasks, and report outcomes — without a human initiating each step. CBIZ's use of Foundry signals that this category of tool is moving from enterprise experimentation to professional services deployment at scale.

What is the first practical step a small firm should take toward an agent-native model?

Map one workflow that costs 3+ hours per week in coordination time — status updates, data gathering, scheduling, report generation. That workflow is your first agent candidate. Before building anything, write down every step the workflow requires: what data is needed, where it lives, who does each step, and what the output looks like. That documentation is the input your first automation needs. Most small firms skip this step and wonder why their AI tools underperform.

CBIZ Agent-Native AI Model for Small Firms

On April 21, 2026, CBIZ — one of the ten largest professional services firms in the United States, with 9,500+ employees — announced something that most observers filed under "large company news" and moved on.

They shouldn't have.

CBIZ announced an expansion of its Microsoft partnership to build an explicitly agent-native operating platform. Not "we're rolling out Copilot." Not "we're piloting AI in one department." They declared their intent to rebuild how the firm operates around AI agents that, in the words of CBIZ VP David Fisher, "act, not just answer" — gathering inputs, triggering tasks, and orchestrating multi-step processes end to end.

This is the first public declaration by a top-10 professional services firm that agent-native is the target architecture. And the blueprint is visible.

What "Agent-Native" Actually Means (and Why the Distinction Matters)

Most professional services firms that have adopted AI are tool-layer firms. They've added AI to their existing workflows. A paralegal uses ChatGPT to summarize a deposition. An accountant uses Copilot to draft a client email. An operations manager asks Claude to analyze a data set.

These are real productivity gains. They're also the floor, not the ceiling.

An agent-native firm is built differently. The workflows themselves are designed around what AI agents can do — not what humans did before AI existed. The difference looks like this:

Tool-layer approach: Your project manager reviews project status across five client engagements every Monday morning by logging into each system, pulling updates, and assembling a summary. With AI, she uses Copilot to help write the summary faster.

Agent-native approach: An agent monitors all five client engagements continuously, identifies status deviations from project milestones, aggregates the relevant updates, flags the two items needing human judgment, and delivers a decision-ready briefing to your project manager Monday morning. She spends 20 minutes instead of two hours — on the items that actually need her.

The tool-layer version saves 30 minutes. The agent-native version recovers 100 minutes and redirects it to client work.

CBIZ is building for the second version, firm-wide.

The 3-Layer Architecture Behind the CBIZ Model

The CBIZ announcement isn't just a vendor partnership story. It describes a specific architecture — three layers, each doing a distinct job — that any professional services firm can replicate at their scale.

Layer 1: The Foundation (Microsoft 365 Copilot)

This is the daily productivity layer. Every employee using AI-assisted tools for the tasks they do every day: drafting, summarizing, researching, analyzing. CBIZ is rolling out M365 Copilot to all 9,500+ team members, with training paths built for each role.

For a smaller firm, the foundation layer doesn't require an enterprise license. It's the decision to get every team member — not just partners, not just senior staff — using AI tools consistently for routine work. The Karbon 2026 State of AI in Accounting report found that firms with 21–50 employees are the most confident AI adopters for the third year running, outperforming larger firms. The reason: full-firm deployment is achievable at that size. At 10 or 15 employees, you can make the foundation layer happen in a week.

The foundation layer generates two things: productivity gains (immediate) and the usage data you need to identify which workflows to automate next (strategic).

Layer 2: The Automation Layer (Copilot Studio)

This is where firm-specific agents get built. Copilot Studio lets firms create AI agents that execute specific workflows — not general assistants, but purpose-built agents for your firm's actual operations.

A CBIZ agent built in Copilot Studio might handle: client onboarding intake, pulling information from a prospect's financial statements and populating a project scope template. Or weekly status reporting across all active client engagements. Or internal knowledge base queries that route specific questions to the right team member based on expertise tags.

For smaller firms, the equivalent tools are Copilot Studio (if you're already in M365), Zapier, or Make. The key is building agents around the workflows you've already mapped in Layer 1. The firms that stall at this layer typically try to build agents before they've documented what the workflow actually requires.

Practical starting point for a 15-person consulting firm: Pick one recurring, structured task that requires gathering information from 2+ places — weekly utilization reports, client status updates, invoice preparation. Document every step of that task in writing. Then build an agent for that one task. One agent, running reliably, is worth more than five partially-built workflows.

Layer 3: The Orchestration Layer (Microsoft Foundry)

This is the layer most small firms don't reach — and it's where the compounding return lives.

Microsoft Foundry (Azure AI Foundry) is the infrastructure for multi-step workflows that cross systems and departments. It's the engine that lets an agent in Layer 2 trigger actions in external systems, receive outputs, make conditional decisions, and route results — without a human in the loop at each step.

For CBIZ, Foundry is what turns individual Copilot Studio agents into an interconnected operating system. An agent doesn't just gather client status — it updates the project management system, flags the billing exception, and alerts the engagement manager. The steps are connected and automated.

For smaller firms, you don't need Azure Foundry at 15 employees. But the principle applies. A workflow that connects your CRM data → your project status → your time tracking → your invoicing system represents the same orchestration logic at a smaller scale. The tools are Make, Zapier, or n8n — the principle is the same.

Most small firms who have "tried AI and not gotten much from it" have built Layer 1 partially and never touched Layers 2 or 3. That's like buying a restaurant's kitchen equipment and using only the microwave.

Why This Is Different From Every Other AI Announcement This Year

Every week, a software company announces an AI feature. Most of them are Layer 1 improvements: better Copilot features, smarter autocomplete, faster document generation.

The CBIZ announcement is different because it's an operating model declaration, not a product feature. CBIZ is committing to redesigning how their firm runs — not just adding AI to the firm that already exists.

David Fisher's framing makes the business case explicit: this is about the ability to "expand value delivery, enter new markets, and deliver deeper insights." Not "reduce costs." Not "automate administrative tasks." Enter new markets.

That's an AI-native growth strategy, not an efficiency play. And it signals where the competitive floor is moving for professional services firms.

The firms that implement Layers 1, 2, and 3 over the next 18 months will be able to deliver more complex client work, to more clients, with the same headcount — or take on more clients without proportional headcount growth. The firms that stay at Layer 1 will be competing on price and availability in a market where AI-native competitors are available around the clock.

What to Do This Week

You are not CBIZ. You don't need 9,500 employees to move.

Here's the 3-step sequence that works at your scale:

Step 1 — Complete Layer 1. Get your whole team — everyone — using at least one AI tool for at least one recurring task. Not just the partners. Not just the tech-forward people. The assistant. The operations manager. The junior staff. Full-firm usage is the foundation.

Step 2 — Map one Layer 2 workflow. Choose the single recurring task that costs the most coordination time each week. Write down every step, every data source, every input, every output. That documentation is your first agent brief. Build one agent for that workflow. Use Copilot Studio, Zapier, or Make — whichever fits your current stack.

Step 3 — Connect two systems. Pick two tools your team uses every day that share data but require manual transfer: your CRM and your project tracker. Your time tracking and your invoicing. Make the connection automatic. That connection is your first orchestration layer.

CBIZ announced their architecture on April 21. The firms that study the blueprint and start building this week will have a meaningful head start by the time their clients start asking about it.

The Crossing Report covers what professional services firm owners need to know about AI — every Monday at 6 AM EST. Subscribe here.

CBIZ Just Built What Every Professional Services Firm Needs — Here's the 3-Step Framework Behind It