The number has been in research reports for two years. McKinsey first estimated it in 2023, then confirmed and raised it in November 2025: AI can automate work activities that account for 60 to 70 percent of what employees do every day. Not by 2030. Not in theory. With tools that exist right now.
For most B2B leaders, that number produces one of two reactions. Either it feels too large to be credible, or it feels overwhelming to act on. Both reactions lead to the same place: doing nothing, while a smaller group of competitors quietly redesigns their operations around it.
This article breaks down what the 60–70% actually covers, which specific AI features are driving it, which business workflows are most ready to automate, and — most practically — how B2B companies are building the case and getting started in 2026.
What does "60–70% of workflows" actually mean — and what does it not mean?
McKinsey's research is widely cited and widely misunderstood. The 60–70% figure refers to work activities — specific tasks within jobs — not to entire roles or headcounts. This is an important distinction that changes how businesses should respond to it.
The research asks: for each individual activity inside a job, could a software agent or robot perform that activity at or above human level using tools available today? The answer is yes for activities covering 60–70% of total working hours. That does not mean 60–70% of jobs disappear. It means that most jobs contain a large portion of tasks that AI can now handle — freeing the humans in those roles to spend more time on the work that AI cannot do: judgment, relationship management, creative problem-solving, and accountability.
"Tasks occupying more than half of current work hours could potentially be automated, primarily by agents. Yet that does not mean half of all jobs would disappear — many would change as specific tasks are automated, shifting what people do rather than eliminating the work itself."
The practical implication for B2B companies is this: every business function contains a set of repeatable, data-handling tasks that are ripe for automation — and a separate set of human-judgment tasks that are not. The companies winning in 2026 are not replacing their teams. They are identifying which tasks belong in which category and rebuilding workflows accordingly.
Which business workflows have the highest automation potential in 2026?
Not all workflows automate equally. The highest-potential workflows share three characteristics: they involve consistent, structured inputs; they produce a predictable output; and they do not require emotional intelligence, physical presence, or high-stakes judgment. Below is how major business functions rank on automation potential, based on McKinsey's task-level analysis and Cflow's January 2026 workflow automation statistics.
The pattern is consistent with McKinsey's finding that digital and information-processing tasks carry the highest automation exposure, while interpersonal, coaching, and leadership-oriented skills are least affected. This is why finance and customer support rank highest — both involve high volumes of structured data handling — while executive strategy ranks lowest.
Which AI features are actually driving workflow automation in 2026?
The jump from 50% to 60–70% in automation potential is almost entirely explained by one development: generative AI's ability to understand and generate natural language. Roughly 25% of all work activity requires natural language processing — reading documents, drafting communications, interpreting unstructured data. Traditional automation could not touch this. Modern AI can.
Here are the six AI features most actively deployed in B2B workflow automation right now:
AI Agents
Autonomous systems that can plan, execute multi-step tasks, and hand off work without human prompting. Used for lead qualification, meeting scheduling, customer query resolution, and code migration. McKinsey found a global tech company deployed agents to qualify sales leads, freeing specialists to spend more time negotiating and building client relationships.
Deployed nowIntelligent Document Processing
AI that reads, classifies, and extracts data from unstructured documents — invoices, contracts, clinical reports, compliance filings. A pharmaceutical company McKinsey studied reduced human touch time on first-draft clinical reports by nearly 60% and cut error rates by around 50% using this approach.
Deployed nowConversational AI & Support Automation
AI systems that handle customer queries across chat, email, and voice — understanding intent and sentiment, not just keywords. A large utility in McKinsey's case studies cut average cost per customer call by approximately 50% using conversational agents while routing complex issues to human agents.
Deployed nowIntelligent Process Automation (IPA)
The combination of traditional RPA with AI decision-making — capable of handling exceptions, learning from data, and adapting outputs dynamically. The IPA market is projected to grow from $16 billion in 2024 to over $18 billion in 2025, reflecting 12.9% CAGR as enterprises move beyond rigid rule-based automation.
Deployed nowPredictive AI for Operations
Forecasting models embedded into supply chain, inventory, demand planning, and risk management workflows. Retailers using AI-powered demand forecasting automatically trigger supplier reorders when inventory thresholds are met — eliminating manual monitoring entirely from that workflow.
Rapidly expandingNo-Code AI Workflow Builders
Platforms that allow non-technical business users to build and configure AI-powered workflows without writing code. Salesforce Einstein Copilot, for example, proactively recommends workflow steps, summarises CRM data, and initiates follow-ups based on natural language instructions — no developers required for standard configurations.
Rapidly expandingWhy are only 13% of companies actually automating at scale — and what's holding the rest back?
If 88% of organizations use AI in at least one function, but only 13% are implementing intelligent automation at scale, something significant is happening between adoption and impact. The gap is not primarily technical. Most of the barriers are organizational.
- Workflow redesign avoidance. Companies deploy AI on top of existing processes rather than rebuilding those processes around AI. McKinsey found that maximum value comes only when AI agents are embedded in redesigned workflows — not bolted onto old ones.
- No dedicated product owner for automation. Pilots succeed when one named person owns the process, monitors quality, and has authority to adapt the workflow. Without this, pilots stall after launch.
- AI skills gap. Deloitte's 2026 enterprise AI report identified insufficient worker skills as the single biggest barrier to integrating AI into existing workflows — ahead of technology cost and data quality.
- Shadow AI fragmentation. Teams deploy AI tools outside enterprise governance — creating security exposure, data inconsistency, and workflows that cannot scale or connect with each other.
- Measuring the wrong things. Companies track "AI features activated" rather than "workflow hours reclaimed" or "error rates reduced." Without outcome metrics, it is impossible to justify scaling or to identify what is actually working.
PwC's 2026 AI predictions research describes what separates companies that scale from those that stall: the most successful businesses build a centralized "AI studio" — a team with reusable components, testing frameworks, deployment protocols, and skilled people — rather than running isolated department-level pilots. This structure links business goals to AI capabilities and surfaces the highest-ROI use cases systematically.
Find verified AI development agencies that have delivered workflow automation for B2B companies like yours
TechRadiant verifies AI agencies on domain track record, real client outcomes, and team depth — not just portfolio looks. Share your automation challenge and get a curated shortlist in 48 hours.
How should B2B companies start automating workflows with AI in 2026?
The companies capturing the most value from AI automation are not the ones with the largest AI budgets — they are the ones with the clearest starting point. The sequence below reflects how the most successful B2B automation programs have been structured, based on PwC, Deloitte, and McKinsey case data.
Audit your highest-volume, most repeatable workflows first
Map which workflows consume the most human hours on tasks that are rule-based, data-driven, and consistent in their inputs. Finance, customer support, and sales operations consistently yield the fastest returns. Do not start with complex, judgment-heavy processes — start where volume is high and variance is low.
Separate automatable tasks from judgment tasks within each workflow
The 60–70% figure applies to task-level analysis, not whole workflows. Within any given process, some steps are fully automatable today, some are partially automatable, and some require human judgment. Map each step and assign it a category before selecting technology — this prevents over-automation (removing human oversight where it is still needed) and under-automation (using AI only for peripheral tasks).
Pilot one workflow with a named owner and clear outcome metrics
Choose one workflow. Assign a product owner who attends every review and has authority to adapt the implementation. Define two or three hard metrics before launch — hours reclaimed per week, error rate reduction, cost per transaction — so success is measurable, not subjective. Deloitte found that companies with concrete outcome metrics for AI are significantly more likely to scale their programs.
Build a reusable component framework, not a one-off tool
PwC's research shows that companies which scale AI automation successfully build centralized "AI studios" with reusable components — authentication, data connectors, prompt templates, monitoring dashboards — that can be adapted for new workflows without rebuilding from scratch. Companies that treat each automation as a one-off project spend 3–4 times more to achieve the same output as those building reusable infrastructure.
Hire or partner with a team that has domain-specific automation experience
Generic AI development agencies can build tools. Domain-specific agencies can redesign your workflows. The distinction matters enormously — a team that has automated finance operations for three B2B SaaS companies understands the edge cases, integration requirements, and compliance constraints that a general-purpose developer would encounter for the first time in your project. TechRadiant's verified AI agency reports filter by industry and use-case track record precisely for this reason.
What is the real ROI of AI workflow automation — and which industries are leading in 2026?
The economic case for AI workflow automation is no longer theoretical. McKinsey's November 2025 research projects that AI-powered agents and robots could generate approximately $2.9 trillion in US economic value per year by 2030 — but only for organizations that redesign workflows around human-AI collaboration rather than isolated task automation.
| Industry | Primary automation use case | Documented outcome | Automation readiness |
|---|---|---|---|
| Financial services | Fraud detection, credit assessment, customer onboarding, compliance reporting | ~50% reduction in cost per customer interaction (McKinsey utility case study) | Very high |
| Healthcare | Clinical documentation, patient eligibility, care prioritisation | ~60% reduction in human touch time on clinical report drafts (McKinsey pharma case) | Very high |
| Manufacturing | Quality control via computer vision, predictive maintenance, supply chain alerts | Real-time defect detection, significant reduction in unplanned downtime | High |
| Retail & e-commerce | Demand forecasting, inventory reordering, personalised CX at scale | Automated reorder triggers, 7–12% projected revenue uplift from AI-led sales (McKinsey) | High |
| IT & software | Code modernisation, incident response, documentation generation | Up to 50% reduction in human hours for code migration (McKinsey bank case) | High |
| Professional services | Contract review, research synthesis, proposal generation, billing | Significant hours reclaimed on document-heavy workflows; still early in scaled deployment | Moderate |
The data across industries points to the same pattern McKinsey identified in its case studies: the highest returns come when humans move from execution to orchestration — managing AI systems, reviewing outputs, and applying judgment where it adds the most value. Companies that treat AI as a cost-cutting tool typically achieve lower returns than those that treat it as a capacity multiplier.
"In every successful case, the human doesn't do less work — they do higher-level work. As AI agents take over routine execution, the definition of 'management' is being rewritten."
