Automating without diagnosis costs more than doing nothing at all because a poorly executed project doesn't just fail; it consumes budget, team time, and internal credibility that the next automation attempt will need to recover. According to Gartner, at least 50% of generative AI projects are abandoned after the proof of concept due to poor data quality, inadequate risk control, escalating costs, or a lack of clarity on business value.
The rush to automate is precisely what produces this pattern. Companies see competitors talking about AI, feel pressured to "do something," and jump straight to choosing a tool without first understanding which process is truly worth automating first.
What a Diagnosis Reveals Before Any Code Is Written
Diagnosis, here, isn't expensive consulting jargon that delays the project. It's the process of understanding how information truly flows within the business before deciding what to automate, and it typically takes weeks, not months.
As Filipe Osanai, co-founder of Draivv, summarizes, the starting point is always to understand the business before the tool: talk to departments, understand how information flows, map out workflows at a macro level, and only then build an impact matrix based on the perception of those who live the process every day. It's not about automating just anything, because the business is at the center of the result, not the chosen technology.
Where the Process Truly Stalls (Not Always Where the Company Thinks)
The perception of those operating day-to-day often reveals a different bottleneck than what leadership imagines. A sales director might think the problem is lead response speed, while the sales team knows the real delay is in internal proposal approval, which passes through three people before going out. Without this direct conversation with those who execute, automation solves the wrong symptom, no matter how sophisticated the chosen technology. This type of hidden bottleneck is precisely what becomes visible when marketing and sales stop operating as isolated departments, a topic explored in depth in the article on marketing and sales as separate departments.
How Much Each Bottleneck Costs Today, in Numbers
Without measuring the cost of each bottleneck, in team time, delivery quality, or lost revenue, it's impossible to know if it's worth solving. A process that consumes two hours per week from an analyst costs much less to automate than a process that delays contract closing by weeks. Putting a number on each bottleneck is what transforms "priority guesswork" into a defensible decision.
Prioritization by Impact, Not by Trend
Based on the bottleneck map with estimated costs, prioritization begins: what has the highest return comes first, not what seems most interesting to showcase internally. Customer service automation might seem more "visible" than financial reconciliation automation, but if the latter saves more hours per week, that's what should come first.
What Happens When This Step Is Skipped
According to the latest McKinsey research, nearly nine out of ten organizations already use AI regularly in at least one function, but only 37% can attribute any impact on EBIT to AI use. The gap between adoption and financial results is precisely the space that diagnosis should fill.
Projects that skip this step often follow a recognizable pattern: they work well in demonstrations with example data, and then stall when they go into production with the company's real data, because they were built on a process assumption, not on the actual process, full of exceptions and edge cases that no one documented.
Another common pattern is the project that works technically but no one uses, because it was designed without understanding how the team actually works today. The tool becomes underutilized, the investment doesn't pay off, and the company concludes that "AI doesn't work for our business," when in reality, the process needed to be understood before building the solution. This same risk appears when automation involves connecting systems that don't currently communicate, as detailed in the article on connecting CRM and ERP with AI agents without losing context.
How to Know if Your Company Has Enough Clarity to Jump Straight to Execution
It's worth asking four questions before contracting any automation tool or project. Is there a clear measurement of where the process stalls today, with data, not opinion? Is there an understanding of how much this bottleneck costs in time, quality, or revenue? Is priority defined by measured impact, not by what seems more modern or easier to present internally? Did the team that will operate the automated process participate in the design, or was the decision made solely by leadership?
Get the next articles by email
New content from Draivv, straight to your inbox. No spam.
If the answer to any of these questions is no, the diagnosis hasn't truly happened, even if the company already has a clear idea of what it "should" automate. This intuitive clarity, without cost mapping and validation with those who operate, is exactly the type of assumption that generates a project that works in the demo and stalls in production.
How to Apply This in Your Operation
The first practical step isn't to choose a tool; it's to understand the business before the technology: talk to departments, map how information flows today, and build an impact matrix based on the perception of those who live the process, following the same logic described in the bottlenecks above.
Draivv's AI for Business Diagnosis applies exactly this process before any tool or automation recommendation, prioritizing by measured impact instead of market trends.
Learn more about Draivv!
Frequently Asked Questions
How long does a diagnosis take before automating?
A well-executed diagnosis typically takes a few weeks, not months, including conversations with involved departments, bottleneck mapping, and prioritization by impact. The time varies with the number of processes and areas mapped, but rarely exceeds one month in medium-sized operations.
Is the diagnosis paid for even if an automation project isn't closed afterward?
It depends on the provider. It's always worth asking directly if the diagnosis fee is deducted from a future project, as this practice exists in part of the market and reduces the risk of paying twice for the same stage.
How do I know if my company is ready to jump straight to execution?
If there is already clear measurement of where the process stalls, how much it costs, and which initiative has the greatest impact, with validation from those who operate the process daily, the company has, in practice, performed its own diagnosis. Otherwise, the previous step is worthwhile before contracting any automation.
Sources: Gartner, Why 50% of GenAI Projects Fail — And How to Beat the Odds, January 26, 2026. McKinsey & Company, The State of AI in 2026: On the Road to ROI, August 25, 2026 (survey of 1,719 participants in 97 countries).
