Workflow Automation Potential: What the Data Shows

Workflow automation potential measures the share of current work activities that AI agents and software can handle without human execution. McKinsey's November 2025 Global Institute analysis put that ceiling at 57% of US work hours — up from roughly 30% in 2023. Most service businesses are capturing 10–15% of that potential. The gap is not a technology gap. It is a process design gap.

In November 2025, McKinsey's Global Institute updated its automation potential estimates for the US economy. The revised figure: 57% of US work hours are already automatable using current, demonstrated technologies. That number describes what AI agents, software automation, and robotics can handle today — not in a projected future.

Most service businesses are running at 10–15% of that ceiling. The gap is not explained by missing tools. The tools that could close it are commercially available. The gap is explained by process architecture — the way work is sequenced and assigned before any tool touches it.

What the 57% figure actually measures

The McKinsey Global Institute's November 2025 report measured activities, not jobs or roles. Researchers assessed the activities that make up each occupation and evaluated what fraction of those activities current AI and automation technology could handle. The ceiling landed at 57% of US work hours.[¹]

That figure nearly doubled from the roughly 30% McKinsey estimated in 2023. The increase reflects two years of AI agent deployment across commercial software — tools that now handle sequential tasks, interpret unstructured data, and execute across connected systems without human direction at each step.

The 57% does not describe any single business. Service businesses — consulting, recruiting, marketing, professional services — sit above that average. Service work is information-dense and repetitive at the task level, even when the overall workflow appears complex. Scheduling, follow-up, data entry, report generation, client status updates, invoice processing: these are the activities that AI agents handle today.

McKinsey's broader projection: AI-powered agents and robots could generate $2.9 trillion in annual US economic value by 2030, and could put 40% of jobs into highly automatable categories if organizations redesign their workflows around automated systems.[¹] The qualifier in that last sentence — "redesign their workflows" — is the difference between the 57% ceiling and the 10–15% most businesses are actually capturing.

Which tasks carry the highest automation potential in a service business

Automation potential concentrates in five task categories. Each has a distinct ceiling based on how much human judgment the task actually requires.

Data collection and entry carries the highest ceiling — above 90% in most service contexts. Pulling data from one system and populating another, logging activity records, updating contact fields after a call: these tasks require no judgment. They require information retrieval and execution. An AI agent completes them faster and with fewer errors than a human working at volume.

Scheduling and coordination runs at 80% automation potential for most service businesses. An agent reads calendar availability, identifies conflicts, drafts coordination messages, and sends confirmations. Edge cases — client preference changes, double-booking resolution, complex time zone dependencies — require human judgment roughly 20% of the time.

Report generation and compilation runs at 75–80% automation potential. A weekly client status report drawing on project management software, time tracking, and communication logs can be assembled by an agent. The remaining fraction — interpreting anomalies, framing the narrative, deciding what requires escalation — stays with the human reviewer.

Client communication — status updates, follow-up sequences, routine inquiry responses — runs at 60–70% automation potential. The agent drafts from context; the human reviews and approves before anything sends. For high-volume, low-variation communication types (intake confirmations, document request follow-ups, payment reminders), the fraction runs higher. For sensitive relationship conversations, the human writes.

Analysis, judgment, and strategy carries the lowest automation potential — below 15–20% in most service firms. Evaluating which client to deprioritize, reading the subtext of a difficult email, assessing whether a candidate is the right fit: these are not automatable with current technology.

Why most businesses stop at 10–15% of the potential

McKinsey's research found that fewer than 20% of organizations have scaled automation across multiple parts of their business — despite virtually all of them having access to the tools that would allow it.[¹] The explanation is process architecture, not tool availability.

Automation tools get added to workflows designed for human execution. A recruiter who already does manual outreach gets an AI assistant that drafts emails. The recruiter still reviews, edits, and sends each email in the same sequence. The tool changes. The process does not. The gain is typically 3–5 hours per week per person — because the process structure was not redesigned, only assisted.

Formstack's research found that teams save an average of 17 hours per week when automation replaces process steps rather than assisting humans doing the same steps.[²] The 14-hour difference between the overlay approach (3–5 hours saved) and the redesign approach (17 hours saved) comes entirely from process architecture, not tool selection.

McKinsey's finding that most AI adopters are stuck in what the research calls "pilot purgatory" — fewer than one in three have scaled beyond a single pilot — reflects this pattern directly. The pilot works. The team then plugs the capability into the existing workflow rather than rebuilding the workflow around it. The pilot's gains do not compound. The ceiling stays far away.

What workflow redesign means in a service firm

Workflow redesign is a process documentation and sequencing project. The technology runs against the resulting process — it does not replace the project itself.

For a recruiting firm, redesigning around automation starts with the trigger events that initiate work. A new job requirement arrives. A candidate application is received. A status update is due. From each trigger, every step to the outcome gets listed. Each step gets a binary question: does this require human judgment, or does it require information retrieval, formatting, and transmission? Steps in the second category go to the agent. Steps in the first stay with the recruiter.

The result is a workflow where the agent holds the coordination thread. A new candidate application triggers the agent to collect required information, run an initial screen against the job requirements, pull comparison data from the CRM, and surface a structured brief for the recruiter. The recruiter evaluates the candidate. The agent handled the logistics of getting that brief ready.

For a professional services firm, the equivalent redesign applies to client reporting. Instead of a consultant assembling a status report from four systems every Friday afternoon, the agent collects and formats data throughout the week, surfaces open items and anomalies, and delivers a draft Thursday evening. The consultant reviews and adds judgment in fifteen minutes. The agent handled the other three hours.

The ceiling is 22 hours per person per week. Most businesses are capturing fewer than four.

The organizations that Formstack measured at 17 hours saved per week had done this redesign work first. They mapped the workflows, identified the agent-eligible steps, and built the automation against the redesigned process. The tools were the same tools available to every other business. The process architecture was different.

How to calculate your automation ceiling

The ceiling for any specific business is calculable by walking through the actual workflows step by step.

Start with the three to five highest-volume, most repeatable workflows — typically client communication, data collection and entry, scheduling, and status reporting. For each workflow:

A typical service business running this exercise across its three highest-volume workflows finds 12–20 hours per person per week in agent-eligible steps. That range is consistent with the 17-hour Formstack benchmark and with McKinsey's estimate that 60% of employees could save 30% or more of their time with automation applied at the process level.[¹][²]

The calculation requires documented processes, not sophisticated tooling. An agent cannot replace a step that has not been written down. That documentation work is itself the prerequisite for any implementation — and the reason businesses with written processes deploy faster and capture more of the potential than those starting from undocumented workflows.

For related context on scoping your first implementation, see which workflows to automate first and what AI agent implementation costs for a small business.

Frequently asked questions

What percentage of business tasks can be automated? McKinsey's November 2025 Global Institute analysis found that 57% of US work hours are automatable using current, demonstrated technologies — up from approximately 30% in 2023. For service businesses specifically, tasks in scheduling, data entry, report generation, and routine client communication carry the highest automation potential, typically above 70–80% within those task categories.

Why do most businesses only capture a small fraction of automation potential? Most businesses add automation tools to workflows designed for human execution. Layering a tool onto an existing process returns 10–15% of the automation ceiling. Organizations capturing 40–57% of potential restructured their workflows so agents handle entire steps, not just assist humans doing the same steps. McKinsey found fewer than 20% of organizations have scaled automation across multiple parts of the business.

How do I calculate the automation potential for my business workflows? Map your highest-volume repeatable workflows step by step. For each step, determine whether it requires human judgment or information retrieval and execution. Steps in the second category are agent-eligible. Count the weekly hours spent on agent-eligible steps across the team — that total is your automation ceiling. A typical service business finds 12–20 hours per person per week in eligible steps across three to five key workflows.

What is the difference between adding automation tools and redesigning for automation? Adding tools to an existing workflow means a human still performs the same sequence of steps, aided by the tool at certain points. Savings are typically 3–5 hours per week per person. Redesigning the workflow means agents own entire steps and the human enters only at judgment points. Formstack's data on organizations saving 17 hours per week comes from this second approach — workflows where automation replaced steps, not assisted humans doing the same steps.

Notes

1. McKinsey Global Institute, "Agents, robots, and us: Skill partnerships in the age of AI," November 2025. https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-skill-partnerships-in-the-age-of-ai
2. Formstack, "Workflow Automation Statistics You Need to Know." https://www.formstack.com/blog/workflow-automation-statistics