Generative AI projects are repeating the mistakes of ERP and CRM. Clear ROI metrics are the difference between failure and impact.
Headlines warn that 95% of GenAI projects fail. MIT’s State of AI in Business 2025 report confirms the number. What is less often discussed is why these projects fail, and more importantly, how success should be measured.
Too many organizations launch GenAI initiatives without clear ROI metrics. They repeat the same mistakes that plagued ERP and CRM rollouts in earlier decades. Budgets are set, vendors are chosen, and pilots are launched, but data is messy, processes are poorly mapped, and outcomes are vaguely defined. The result is failure rates that analysts repackage as the “trough of disillusionment.”
At Brookey & Company, we believe ROI is not abstract. It must be measurable across financial, operational, and organizational dimensions. This is the foundation of our Expert-Driven Loop (EDL) model and our OrgCore™ assessments.
Why ROI Definitions Fail Today
Executives often ask, “What does success look like for GenAI?” Many projects default to answers such as “efficiency,” “productivity,” or “transformation.” These words are vague. They do not tie to a P&L line item, nor do they withstand scrutiny from a board.
Human-in-the-Loop (HITL) frameworks, and even their more sophisticated cousin, Expert-in-the-Loop (EITL), make the problem worse. Both rely on oversight after the fact. Errors are corrected only once they are visible, long after value has been lost. This reactive posture inflates costs and delays adoption.
True ROI requires clarity at the beginning. Metrics must be defined, tested, and aligned to business outcomes before pilot projects convert to enterprise scale.
The Brookey & Company ROI Framework
Our approach defines ROI across three categories, each with specific metrics:
Financial ROI: measurable P&L impact, cost avoidance, and reduced external spend.
Operational ROI: cycle-time compression, error-rate reduction, and throughput gains.
Organizational ROI: clarity of decision rights, adoption by domain experts, and workforce capability uplift.
This framework ensures ROI is engineered into projects. It is how EDL, Expert-Driven Loop, avoids the failures of HITL patchwork.
ROI Across Sectors
Financial Services
Compliance consumes 10–15 percent of operating expenses for large banks. Even a five percent efficiency gain translates to millions in savings. ROI comes from lower compliance costs, faster risk checks, and reduced reliance on expensive outsourcing.
Healthcare and Pharma
Clinicians spend up to half their time on documentation. Automating transcription, notes, and administrative workflows yields clear ROI: more patient-facing hours, lower error rates, and faster throughput in clinical operations. Financial impact appears in staff utilization and compliance strength.
Consumer and Retail
AI-driven engagement can shorten lead cycles by 20–30 percent. ROI is seen in faster qualification, improved sales conversion, and measurable increases in retention. Unlike traditional CRM deployments, success here is expressed as marginal revenue lift and loyalty metrics, not just transaction counts.
Technology and Media
These industries are at the frontier of disruption. ROI is measured in content throughput, cycle-time compression in software engineering, and reduced ad production costs. For example, an AI system that accelerates content creation by 25 percent delivers immediate savings while enabling faster speed-to-market.
Professional Services
Billable hours are the currency of consulting and legal firms. Automating document review, transcription, and research directly recovers revenue-producing time. ROI is concrete: more hours billed, fewer write-offs, and reduced non-chargeable workload.
Advanced Industries (Manufacturing, Energy, Materials)
ROI is strongest in back-office and operational optimization. Predictive maintenance reduces downtime and extends asset life. Supply-chain automation compresses procurement cycles and reduces reliance on BPO providers. The financial gains outweigh those from customer-facing pilots.
Lessons from Past IT Implementations
GenAI failures are not new. They mirror ERP and CRM disasters that executives have seen before. The patterns are consistent:
Poor data hygiene leading to unreliable outputs
Weak or shifting requirements that undermine scope
Lack of expert involvement in defining requirements
Centralized IT-led rollouts that ignore domain context
ROI not defined clearly at the outset
Over the years, the industry has tried to patch these gaps with new oversight models. Human-in-the-Loop (HITL) was positioned as the solution, followed by Expert-in-the-Loop (EITL). Both improve accountability on paper, but in practice they remain reactive. Errors and inefficiencies are caught only after they have already affected outcomes.
The technology was not the cause of ERP failures, and it is not the cause of GenAI failures. The real cause is weak organizational design, poor requirements discipline, and reactive oversight.
Organizational Design as the Critical Factor
MIT’s report confirms that the barrier to GenAI success is not budget or model quality. It is organizational design. The data is clear:
66 percent of successful deployments came from strategic partnerships where companies co-developed with vendors.
33 percent of successes came from internal builds.
Hybrids lagged, often due to unclear ownership.
The conclusion is unmistakable: companies succeed when they decentralize implementation authority while retaining accountability. They fail when they centralize control and treat GenAI like a commodity IT purchase.
This aligns directly with our OrgCore™ approach. Organizational design determines whether AI crosses the divide or stalls.
The Brookey & Company Approach
Through OrgCore™ assessments and the EDL model, Brookey & Company translates these lessons into practice:
Decision rights and accountability mapping: clarity on where authority must sit.
Build-buy readiness: defining whether a strategic partnership or internal build is the right path.
Outcome instrumentation: linking GenAI to measurable cost pools, revenue lifts, and operational metrics.
Frontline sourcing with executive governance: capturing use cases from managers and prosumers while maintaining board-level accountability.
Learning and orchestration roadmap: evolving point solutions into adaptive, memory-capable workflows.
Conclusion: ROI Is the Difference Between Hype and Clarity
ROI from GenAI is not universal, but it is always measurable when organizations define it clearly, align it to outcomes, and build accountability into the process.
The current “trough of disillusionment” often described by analysts has little to do with the underlying technology. It has everything to do with how organizations approach AI projects. Just as with ERP and CRM, success depends on clean data, mapped processes, and well-defined outcomes. When those fundamentals are missing, failure is inevitable.
At Brookey & Company, we ensure ROI is not left to chance. Through OrgCore™ assessments and our Expert-Driven Loop (EDL) model, organizations build the discipline, clarity, and accountability to cross the GenAI Divide and capture measurable returns without wasted time or overhead.
Frequently Asked Questions
Why do most GenAI projects fail?
Most failures stem from organizational design problems, poor data hygiene, shifting requirements, and reactive oversight. Success requires clear ROI metrics, mapped processes, and defined accountability.
How should executives measure ROI from Generative AI?
ROI should be defined before pilots begin, across financial, operational, and organizational dimensions. That means linking cost avoidance and reduced external spend to P&L impact, tracking cycle time and error rate reductions, and ensuring expert adoption and decision rights clarity.
Does ROI vary by sector?
Yes. Each sector defines ROI differently:
Financial services: compliance cost reduction, faster risk checks.
Healthcare and pharma: documentation efficiency, improved throughput.
Consumer and retail: lead-cycle compression, higher conversion rates, loyalty lift.
Technology and media: content throughput, reduced engineering cycle time.
Professional services: billable hours recovered through automation.
Advanced industries: predictive maintenance savings, supply chain efficiency.
What is the difference between HITL, EITL, and EDL?
Human-in-the-Loop (HITL) and Expert-in-the-Loop (EITL) are reactive oversight models, designed to catch errors after they occur. Expert-Driven Loop (EDL) is proactive. It defines ROI from the outset, embeds domain expertise into the process, and validates outcomes through structured pilots before scaling.
What is OrgCore™?
OrgCore™ is Brookey & Company’s organizational assessment suite. It helps companies map decision rights and accountability, evaluate build-versus-buy readiness, instrument outcomes to measurable cost pools and revenue lifts, and plan a roadmap for orchestration and learning.
Reference: MIT, The GenAI Divide: State of AI in Business 2025, July 2025. Read the full report (PDF).