Gartner recently made a stark prediction: over 40% of agentic AI projects will be canceled by 2027, citing rising costs, governance challenges, and lack of clear ROI.
We agree that many will fail — but not because AI is falling short. The real issue is the wrong AI projects being prioritized, with outdated ROI benchmarks to measure success.
And this shakeout isn’t a setback. Rather, it’s a reset that will speed up the rise of agentic AI that truly delivers value. We’re still in the early days, and the winners will be those who continue to prioritize, iterate, and adapt as the technology matures.
In this article, we’ll cover:
- Why agentic AI projects are faltering
- Why ROI is being measured the wrong way
- Which projects will survive the coming shakeout
- How Trullion is building agentic AI designed to last
We’re entering the “Trough of Disillusionment”
According to Gartner’s Hype Cycle, every emerging technology moves through five phases:
- Innovation Trigger
- Peak of Inflated Expectations
- Trough of Disillusionment
- Slope of Enlightenment
- Plateau of Productivity
Agentic AI is currently at the Peak of Inflated Expectations – and is headed into the Trough of Disillusionment. Expectations are falling fast, and Gartner points to:
- Unclear or intangible ROI
- Governance and compliance risks
- Workflow integration difficulties
Every transformative technology hits this phase. The dot-com crash paved the way for today’s internet-native leaders. The cloud was initially dismissed as “too expensive” before becoming an innovation staple.
Early AI failures aren’t the end. They’re the natural filter toward smarter, sustainable adoption. Rather than fearing this phase, organizations should welcome it. The sooner inflated projects collapse, the faster the industry can redirect energy and investment toward valuable solutions.
Why most agentic AI projects are failing
“Agentic AI” has become a catch-all buzzword. Gartner calls out “agent washing” – rebranding chatbots and RPAs as agents without any meaningful autonomy.
True agentic AI perceives, reasons, and acts semi-autonomously. Agents initiate and complete tasks, not just respond to queries.
So why are so many failing?
- Outdated ROI expectations. Many projects are being judged against narrow cost‑savings metrics – instead of measuring long‑term productivity, accuracy, and compliance benefits.
- Lack of domain expertise. Generic agents often fail in high-accuracy fields like accounting, where nuanced knowledge is required.
- Workflow misalignment. Agents that can’t embed into ERP, audit, or financial systems introduce friction rather than efficiency.
- Overhyped technology. Many tools are marketed as agentic, but lack actual autonomy or business value.
At Trullion, we see these project failures as a necessary correction – and proof that AI must be applied correctly and measured fairly.
Rethinking ROI: The real survival test
Here’s where we part ways with Gartner. Yes, many agentic AI projects are failing because they’re flawed – but many are being written off unfairly, judged by outdated ROI benchmarks.
In the 1910s, factory owners debated whether electricity was cheaper than steam, missing its productivity revolution. In the 2000s, cloud skeptics dismissed it as “too expensive,” only to watch it transform business speed and scale.
AI is facing the same scrutiny. Today’s ROI models fixate on short‑term metrics like headcount reduction or immediate cost savings, overlooking the real value drivers:
- Productivity gains that compound over time
- Accuracy and compliance improvements
- Proactive risk mitigation, through intelligent anomaly detection
One of the biggest traps is waiting for “perfect” ROI proof before taking action. For emerging technologies, the clearest returns often come after a phase of experimentation and iteration.
Cloud, mobile, and even the internet all followed this pattern. AI is no different.
Even if ROI is hard to measure today, your investment pays off in long-term efficiency and readiness – especially with domain-specific, embedded agentic AI. The companies waiting today will be the same ones struggling to catch up once the winners are clear.
Which agentic AI projects will survive?
By 2028, Gartner predicts that:
- 15% of work decisions will be made by agentic AI (up from 0% in 2024)
- 33% of enterprise applications will embed agentic AI (up from <1% in 2024)
The survivors will not be generic agents. And in accounting and finance, specifically, they’ll be purpose-built for high-accuracy and compliance-driven work. Winning projects will:
- Solve defined pain points with domain expertise (e.g. audit prep, lease abstraction)
- Embed into existing workflows, not just serve as bolt-ons
- Prioritize auditability and compliance
- Deliver measurable value in accuracy, time, or cost savings
Visual framework: The AI survivability matrix
Our AI survivability matrix maps which projects are most likely to last:
- X-axis: Domain Specificity. How deeply the AI understands and is tailored to a field like accounting or audit.
- Low specificity: Generic assistants that can operate across many functions but lack deep contextual knowledge.
- High specificity: AI trained on industry-specific data, workflows, and standards (e.g. ASC 842, IFRS 16, SEC rules).
- Y-axis: Workflow Integration. How well the AI embeds into core processes.
- Low integration: Agents that operate in silos or require major process changes to adopt.
- High integration: Agents that seamlessly fit into existing systems, saving time and improving accuracy.
The future of agentic AI in finance and accounting
Finance is well positioned to benefit from agentic AI — especially when tools are built with domain workflows in mind. High-impact use cases on the horizon include:
- Continuous close and real-time reconciliation. Books no longer close once a month. They’re perpetually audit‑ready. Agentic AI reconciles matches contracts, bank statements, and ERP entries as they happen, and surfaces discrepancies before they become problems.
- Smart audit preparation with AI evidence gathering. Audits shift from reactive marathons to proactive check‑ins. AI agents compile supporting documents, link them to ledger entries, generate PBC lists, and package evidence, cutting weeks off audit prep time.
- Dynamic lease and contract accounting. No more manual recalculations when agreements change. Agentic AI interprets leases and revenue contracts, extracts key terms, and recalculates journal entries automatically — ensuring compliance with standards like ASC 842 and IFRS 16.
- Standards enforcement with adaptive intelligence. Agentic AI automatically applies the latest standards and internal policies, minimizing the risk of misapplication or audit findings.
- Conversational AI, built for accounting and audit. AI agents have a dedicated understanding of your accounting standards and financial data. Imagine a colleague who knows every contract, policy, and journal entry — and can answer your question in seconds, always citing their sources.
Trullion’s approach: AI that lasts
We’ve always believed that AI should enhance human expertise, not replace it. Precision and compliance are critical, and so is proper governance. We built our platform around four core pillars:
- Domain specialization. AI has deep knowledge in accounting, audit, and compliance. It understands operational nuances.
- Traceability-first design. Every data point extracted is linked back to its source, ensuring full auditability and transparency.
- Human-in-the-loop review. Professional human judgment always happens where it matters.
- Continuous investment. We actively expand ERP and audit system integrations to enhance workflow impact.
The takeaway: Don’t get left behind
Just like no modern company questions the ROI of internet access, the next generation of enterprises won’t question AI usage – they’ll compete on how well they use it.
Embrace the shakeout, let failed experiments fall away, and double down on the agentic AI. The winners will be those who act early, focus on domain-specific solutions, and embed AI into workflows with full transparency. That’s exactly what we’re building at Trullion.
Curious how our agentic AI can help you deliver audit-ready insights, fast? Meet Trulli, your always-on financial assistant.