In this edition of Ashling Unfiltered, Jeffrey Brown, Managing Director of Advisory at Ashling, puts Observatory Circle Advisors' "Tools, Tools, Tools" under the microscope. He pull the headline claim, checks it against field reality, and ends with one move you can make this quarter. Let's dive in.
Enterprises poured billions into Generative AI in 2025, yet the ROI headlines are brutal. MIT's Project NANDA reported that 95% of enterprise AI initiatives have delivered close to zero impact, a number that lit up LinkedIn for months. If that 95% failure stat sounds familiar, it should. Our previous Ashling Unfiltered, “Decoding the GenAI Divide: How to Be the 5% That Ships” scrutinized the report for its ambiguous definition of AI success, self-selected data, and sweeping generalizations.
Observatory Circle echoes a similar point: The ROI is there, but it isn't apparent. Enterprises are measuring it wrong, operating with models that were never designed for AI, selecting the wrong tools for the wrong business impact, and leaking value to disruptors that are experimenting earlier and learning faster.
Observatory Circle gets the big picture right. Significant value is being created by AI; it's just not showing up cleanly on incumbents’ P&L statements, yet. But our market experiences add important nuances to each of their four diagnoses.
We disagree. Traditional ROI metrics such as cost savings remain valid; the real failure is in translating softer metrics into hard-dollar value. Measuring Agentic AI success solely through a traditional ROI lens is like measuring the value of the internet by the number of phone calls it replaced. This metric paints a narrow picture of the total value returned. Take an AI assisted commercial loan underwriter, who can reduce their average handle time per case by 40-50%. While a bank could capture that “time value” via staff reductions, most will take the opportunity to more rapidly grow a profitable commercial loan book, which tends to lead to additional retail and wealth management business.
Strongly agree. This is the most underappreciated dynamic in the market right now. We've seen it play out before with digital-native firms outpacing analog-first incumbents through the 2000s and 2010s. AI-native businesses are running the same playbook. Incumbents who wait for the "mature" use case are already behind. The most vulnerable of the incumbents are the outsourcing firms who are racing to use AI approach before their clients do it themselves.
Strongly agree, and this is where most organizations are wrongly invested. Too much attention and investment go to the AI engine and not enough to the surrounding business system: the process reimagination, data, governance, the digital workers executing tasks, and the people actually driving the process. Getting value from AI is as much a change management project, as it is a technology project. Too many teams try to upgrade their current processes by just adding a new shiny AI tool, instead of reimagining the end-to-end process from scratch. Remember the folly of putting pdf applications onto emerging digital channels?
This is the single biggest driver of why ROI looks so poor, and why CFOs are losing faith in AI initiatives. Many organizations are installing new F1 engines into a five-year-old Honda family car, so teens can get to high school safely. Worse, they're defaulting to a single tool (like an LLM) for an entire process, when a well-composed stack of complimentary tools would perform far better, in terms of speed and accuracy, and cost significantly less. Few organizations are systematically capturing the economies of skill and scope that compound over time from multiple AI initiatives.
There's no shortage of implications from Observatory Circle’s insights. But here are some of the moves that you can make next quarter:
Start where skills bottlenecks meet high-value processes: Loan underwriting, claims assessment, KYC — these are areas where AI-saved time can be redeployed immediately into higher-value, revenue generating work. More rapidly growing firms earn higher valuation multiples.
Audit your tool stack for 'F1 car' problems: For every active AI initiative, ask: is this AI tool doing work a simpler, cheaper tool could handle? Are we paying avoidable costs for deterministic logic? Rebalancing the stack is often the fastest path to positive ROI.
Act like a disruptor to your own business: Evaluate which outsourced or BPO-dependent processes could be rebuilt in-house with a leaner, AI-augmented team. The math often surprises people. At a minimum, you’ll have a clear handle of the value to extract from your BPO partner.
Align your measurement methodology for AI economics: The measurement framework your organization uses signals to the entire organization about what success looks like. And what gets measured, gets done. The inverse is equally true. An AI Center of Excellence (CoE) needs to adjust how value and costs are calculated given the rapidly changing economics of AI. If done right, it will tell a fundamentally different story – there is significant value