3 Ways AI Can Free Organizations from Legacy Workflows

Elena* leads a once-innovative logistics firm we’ve studied that we’ll call Virtal Systems. It’s now struggling to keep pace. “We’re not short of capability,” she explained to us, “we’re weighed down by our own past.” Legacy workflows persist, old assumptions guide decisions, and “the way we’ve always done it” shapes strategy. In a market transformed by real-time data and automation, these habits quietly erode competitiveness. As Virtal illustrates, what often constrains a business’s competitiveness is not what a company lacks, but what it carries forward via corporate memory. Managers cling to entrenched methods that no longer reflect the realities of today’s market. The requirement that an organization moves on from its past and starts afresh is known as “organizational forgetting.” Yet this forgetting is frequently neglected, despite its importance in an environment defined by expanding data infrastructures and the growing influence of AI on organizational decision-making. Here we outline three organizational constraints that arise when executive teams struggle to let go of the past. We also provide examples of how you can employ AI to renew your organization’s practices for an improved competitive future and use AI to make an objective case for change. Performance Metrics that Distort Priorities Clinging to obsolete metrics leads companies to misallocate resources, reward the wrong behaviors, and miss up-to-date performance drivers that lead to organizational success. In the age of “digital dashboards” this danger is compounded as outdated KPIs don’t just linger in reports, they become visualized, cemented, and circulated. This embeds distortion deep into decision-making systems. Take a national U.K. retail chain we studied—we’ll call it Whitford & Co. It’s a member-owned consumer co-operative best known for its food retail stores. Its sales were stagnating, margins were tightening, and customer churn was rising. Despite leadership’s best efforts to improve results via change-management programs, outcomes remained stubbornly unchanged. Out of habit, staff clung to legacy KPIs that they’d become accustomed to. How AI helped: Metrics overhaul initiative A core stumbling block in an organizational forgetting program is complexity. To expose which of its legacy metrics were relevant and which ones weren’t, Whitford & Co deployed two AI-powered analytics platforms, Snowflake Cortex Agents and Microsoft Fabric, to do what human analysts could not, i.e., simultaneously interrogate customer transaction data, digital interaction logs, and operational records. AI processed thousands of data points which exposed that many of the most emphasized metrics had no strong correlation with either customer retention or business profitability. For example, the organization’s primary KPI, in-store sales conversion (percentage of physical store visitors who complete a purchase), reflected an old view of customer behavior. In a world of hybrid customer journeys, this metric captured only the final physical touchpoint, ignoring upstream digital interactions that increasingly shape current purchase decisions. Complexity wasn’t the only issue Whitford & Co faced. Senior management also found that managers didn’t want to let go of the old for the new. With AI, senior managers could make a strong case for retiring KPIs that was not tied to any individual’s or group’s opinion. As Vesna, the director of commercial performance in the retail division, told us: “Employing AI provided an objective, data-backed case for retirement of each redundant KPI. It removed the emotional and political friction that typically derails these exercises.” Over three months, the retail division successfully retired seven of its 12 legacy KPIs, including in-store sales conversion and footfall-to-inquiry ratios (the rate at which in-store visitors converted to service inquiries). In their place the company adopted behaviorally relevant indicators, such as multi-channel path completion rates (the proportion of customers who complete a purchase journey across digital and physical touchpoints combined) and customer effort scores (a measure of how easy customers find it to interact with the business). Business Identities that Confuse the Market Corporate identities and brand positioning that were once assets can become liabilities. Take the U.S.-based software company we studied, which we’ll call FengSys. This company builds data integration and analytics platforms for public and private sector customers. It built its brand around being “data-first” in the early 2010s. Its go-to-market language, sales training, and customer proposals all revolved around this positioning. Over time, though, client expectations changed. Clients were no longer impressed by “data-first.” They had come to take data capability for granted and were now seeking partners who could demonstrate business outcomes—faster decisions, lower operational risk, and measurable return on technology investment. Despite multiple strategy offsites, FengSys’s staff found it difficult to let go of its original identity. The result was that senior management continued to bolt new positioning statements onto outdated ones. For example, newer phrases like “insight-driven transformation” and “AI-ready infrastructure” were layered into sales decks and proposals that still opened with “data-first” language. How AI helped: Strategic language purge FengSys’s organizational forgetting problem required AI to first recognize patterns, specifically, the hidden maze of contradiction embedded across thousands of documents that no human reviewer could reliably detect on budget and at scale. As internal staff were too close to the company’s history to see the incoherence clearly—they had written much of the material themselves and were blind to its poor alignment with the new strategy—the company worked with an external boutique consultancy. This firm deployed a GPT-4-based model configured specifically around the company’s own document library, to scan thousands of pages of sales decks, proposals, and client communications. This meant that the process occurred without the emotional attachment that had paralyzed previous efforts. The model was tasked with detecting contradictory statements, obsolete terminology, and incoherent positioning across the entire document library. For example, it flagged 23 distinct messaging variants in use across the sales team, each a legacy fragment of a different strategic moment in the company’s history. Many of these appeared alongside the dominant “data-first” framing in the same documents. The result? As Leonor, the vice president of marketing explained: “This gave us an objective, evidence-based picture of the problem that no internal advocate could dispute or deflect. When staff saw the contradiction matrix—the visual evidence of how fragmented and incoherent the messaging had become—resistance faded. The problem was no longer a matter of opinion. It was a fact, documented and displayed.” Customers Myths that Warp Strategy A global financial services firm we studied that we’ll call SuboBank clung to the belief that “older customers avoid mobile banking.” This myth originated a decade ago and was embedded in the firm’s training manuals and customer segmentation logics. It also persisted in its product design choices. Even though external data suggested that retirees now constitute one of the fastest-growing groups of mobile app users, the belief that they were technological laggards was hard to shake. SuboBank worked with a single entrenched assumption, confidently held, that had never been systematically tested against real customer behavior. No internal team had the authority or the tools to challenge it. How AI helped: Myth-busting through behavioral evidence To unearth this embedded belief, management configured IBM Watson Analytics to ingest and cross-reference three live data streams simultaneously: 1) in-app behavioral logs tracking how customers of all ages actually used SuboBank’s mobile platform; 2) customer-sentiment data drawn from support interactions, survey responses, and complaint records; and 3) external demographic data on mobile adoption trends among customers aged 60 and over. As Kevin, head of customer insights explained: “We pointed the IBM Watson Analytics engine simultaneously at two things: the firm’s own internal documentation, such as training manuals, product design files, and customer segmentation records, and the live data streams that track the actual choices of customers. The system read the documentation for embedded assumptions about how older customers behaved, then tested those assumptions continuously against the real behavioral evidence.” Wherever actual behavior contradicted what the documentation assumed, the system flagged it. This exercise surfaced several surprising insights. First it demonstrated that customers aged 65 and over were logging into SuboBank’s mobile platform more frequently than any other age cohort—an average of 11 times per month compared to eight times for customers aged 25 to 40. Second, this older cohort submitted the highest volume of support requests, not because they were struggling, but because they were actively seeking to do more with the platform than its limited interface allowed. Third, sentiment analysis of support transcripts revealed that older customers expressed the highest levels of frustration, not with mobile banking as a concept, but with SuboBank’s specific design choices, which had been built around assumptions of their disengagement. These facts made the myth impossible to defend. No internal champion of the old belief could argue against usage logs and sentiment data presented systematically across hundreds of thousands of interactions. As managers and senior leaders reviewed the AI-generated behavioral reports, they made their own judgments about which design changes to prioritize. . . . You’ve probably found that organizational forgetting is hard. That’s because it runs counter to ingrained organizational habits which see leaders attached to metrics that no longer validate performance, teams clinging to language that no longer wins markets, and data that no longer explains outcomes. You can overcome this with AI. What’s more, it can take personal views and emotional hunches out of strategy analysis and present a watertight case for organizational change. Authors’ Note: All individual and company names have been changed in this article to protect confidentiality.

3 Ways AI Can Free Organizations from Legacy Workflows

Summary.

Partner Center