Harvard Business Review published an article on September 14, 2026 that every business owner investing in AI should read. Titled "Stop Automating Old Processes. Design New Ones Instead." by Masha Shunko (University of Washington) and Serguei Netessine (Wharton), the core argument is simple and disruptive: companies are pouring money into AI but failing to generate enterprise-wide returns because they automate individual tasks rather than redesign the workflows that produce business value.
The data backing this argument is sobering. BCG's January 2026 "AI Radar" survey of 640 CEOs across 16 markets found that corporations expect to roughly double AI spending this year, from 0.8% of revenue to approximately 1.7%. Yet PwC's 2026 "Global CEO Survey" reports that only 12% of CEOs see both revenue and cost benefits from their AI investments. McKinsey QuantumBlack's April 2026 analysis goes further: 60% of organizations still see no enterprise-wide EBIT impact, even as nearly eight in 10 use generative AI in at least one business function.
The gap between investment and return is not a technology problem. It is a process design problem.
The Core Insight: Automate Tasks, Fail. Redesign Processes, Win.
The HBR article identifies a pattern we see repeatedly at Vivid Visuals when auditing client operations. A business adopts an AI tool to automate an existing task, say, generating email responses or extracting data from invoices. The task gets faster. But the workflow surrounding it does not change. The email still enters the same inbox, gets triaged by the same human, routes to the same department, and follows the same approval chain. The bottleneck simply moved.
Netessine summarized the argument on LinkedIn: "AI pilots often fail when they automate existing tasks instead of redesigning the workflow first." This is the critical distinction. Task-level automation makes a step faster. Process-level redesign removes steps entirely, changes the order of operations, reassigns decision rights, and creates measurement systems that capture the real value AI delivers.
What the Data Actually Shows
The three studies cited in the HBR article paint a consistent picture of institutional over-investment and under-return:
| Study | Sample | Key Finding |
|---|---|---|
| BCG AI Radar (Jan 2026) | 640 CEOs, 16 markets | AI spend doubling from 0.8% to 1.7% of revenue |
| PwC Global CEO Survey (2026) | CEOs globally | Only 12% report both revenue AND cost benefits |
| McKinsey QuantumBlack (Apr 2026) | Organizations globally | 60% see no enterprise-wide EBIT impact; ~80% use gen AI in at least one function |
The implications for SMBs are twofold. First, if the world's largest enterprises with armies of data scientists and consultants are struggling to show ROI, the bar for smaller organizations is even higher. Second, the reason for failure is replicable and avoidable: automating old processes instead of designing new ones.
A Practical Framework for Redesigning Processes for AI
The HBR article outlines a structured approach to deciding what to redesign and how. We have adapted the core principles into a framework we use with our own clients.
Step 1: Map the End-to-End Workflow Before Touching AI
Before deploying any AI tool, map the full workflow from trigger to outcome. Not the task you want to automate, but the entire process that the task sits inside. If you want to automate invoice processing, trace the path from receipt through approval through payment through reconciliation. You will almost always find that the task you want to automate is one node in a chain with multiple handoffs, delays, and manual checkpoints that AI can eliminate entirely.
Step 2: Identify Which Steps AI Eliminates, Not Just Accelerates
For each step in your mapped workflow, ask: does AI make this faster, or does it make this unnecessary? If the answer is "faster," you are automating a task. If the answer is "unnecessary," you are redesigning a process. The biggest ROI gains come from elimination, not acceleration.
Example: A client wanted to automate their lead qualification process. Their original workflow involved a form, a human review, a data-entry step into the CRM, and an email follow-up. Instead of just automating the email, we rebuilt the workflow: the form feeds directly into an AI agent that qualifies, scores, and routes the lead in real time. Three steps became zero. The human reviewer shifted from data entry to closing.
Step 3: Reassign Decision Rights
When AI handles routine decisions, human decision-makers should move to exception handling and strategic judgment. One of the most common mistakes we see is businesses using AI to draft decisions that humans still rubber-stamp. If the AI is right 95% of the time, redesign the workflow so humans only see the 5% that needs review. If that feels uncomfortable, set a threshold and monitor it. But do not force AI to operate within the decision cadence of a pre-AI workflow.
Step 4: Build Measurement Around the Process, Not the Task
If you measure the time saved on one task, you will see a 20% improvement and declare victory while the overall process is still bottlenecked elsewhere. Measure cycle time, error rate, and cost per outcome at the process level. The HBR article emphasizes that the metrics for evaluating AI should be redesigned as deliberately as the process itself.
Step 5: Start with One Process, Not a Company-Wide Initiative
The BCG data shows that companies are doubling spend, but the PwC and McKinsey data show returns are not following. One reason is that companies launch broad AI initiatives rather than redesigning one process end-to-end, proving the model, and expanding. Pick one workflow that causes recurring pain. Redesign it fully. Measure the result. Use that proof to fund the next one.
The Warning for SMB Owners
If you are a small business owner being sold an AI tool that promises to "automate your workflow," ask a precise question: does this tool automate the tasks in my existing process, or does it let me redesign the process itself? If the vendor cannot answer that question clearly, you are likely buying task automation wrapped in process language.
The HBR article's value is in making this distinction explicit. The investments that generate enterprise-wide returns are not the ones that make individual tasks faster. They are the ones that let you rethink how the work itself is structured. AI is not a faster version of your old tools. It is a reason to redesign the workflow your old tools were built to support.
Sources: Masha Shunko and Serguei Netessine, "Stop Automating Old Processes. Design New Ones Instead." Harvard Business Review, September 14, 2026. BCG AI Radar, January 2026. PwC Global CEO Survey, 2026. McKinsey QuantumBlack analysis, April 2026. Netessine LinkedIn post, September 2026.
Read next from the blog : GEO Becomes Holiday Retail's Fastest-Growing Service: Fiverr's 2026 Index · Constant Contact in Claude: When Your Marketing Tool Lives Inside an AI · AI Bots Are Now Buying on Your Site: What Akamai's Discovery Means for Commerce · The 40% Failure Rate: Why AI Agent Projects Get Canceled — and How SMBs Beat the Odds
