94% of Enterprises Aren’t Seeing AI ROI

The Announcement

Salesforce has launched Agentforce Help Agent, a preconfigured AI service agent deployable across voice, web, portals, and messaging, paired with a pay-per-resolution pricing model that charges customers only when the agent resolves an issue autonomously. Escalations to a human or negative customer feedback are not billed. The move comes directly in response to an admission from Salesforce’s leadership: at the company’s Agentforce World Tour event, President of Enterprise and AI Technology Joe Inzerillo stated that 94% of organizations on the AI journey are not seeing ROI, despite the time and money invested.

Why This Matters for BFSI and Collections

Salesforce backed the launch with its own deployment data: 4.3 million customer inquiries handled through its support portal, with 70% resolved autonomously, alongside a U.K. law enforcement case study where an assistant named Bobbi handles 70% to 75% of non-emergency contacts without human involvement. Among the customer deployments Salesforce highlighted was PenFed Credit Union, giving BFSI operations a direct reference point. For collections and financial services leaders evaluating voice AI, the more significant shift is the pricing logic. Tying cost directly to successful autonomous resolution reframes the vendor conversation from “how much does the platform cost” to “how much does it actually resolve,” which is a more honest question for any BFSI operation trying to justify AI spend to a board that has heard inflated ROI promises before.

What the Numbers Do Not Say Out Loud

A 94% ROI failure rate is an admission that most deployments fail for reasons unrelated to the AI model itself: fragmented data sources, disconnected workflows, and limited agent access to systems that hold the information needed to resolve a case. Salesforce’s fix (pre-built knowledge integrations and single-configuration channel deployment) acknowledges that most of the AI project was never really about the AI, but about the plumbing around it. The 70% resolution figures from Salesforce’s portal and from Bobbi are genuine, but they come from environments where that plumbing was built out first. The pricing model protects against paying for failed resolutions, but it does nothing to protect against the integration and workflow work required to reach a working deployment in the first place.

The Practical Read

For collections and BFSI operations, the lesson is not that outcome-based pricing solves the ROI problem; it is that it exposes where the real risk in any AI deployment sits. It sits not in the model’s capability, but in whether the surrounding data and workflow integration was done properly before go-live. A voicebot handling a payment conversation is only as reliable as the account, case, and compliance data it has access to at the moment of the call. Evaluating a pricing model is the easy part. Evaluating whether a vendor’s implementation genuinely connects to your existing systems, rather than operating on a curated dataset, is the consequential question.

[Read the full report]

Related Post

A Bank Just Drew a Hard Line Between What Its AI Can Say & Decide To Do

The Rollout Commonwealth Bank of Australia has begun testing CommBank Companion, an AI-powered conversational assistant built directly into its mobile app, backed by roughly $1 billion in security investment. Unlike most vendor-led AI launches, this one was designed and built entirely in-house by CommBank’s own engineering and data science teams, and it is being tested […]

Enterprises Are Underestimating Multi-Model AI Failure Rates by 2.25x

A new study evaluating 67 frontier models from 21 providers, including GPT-5.5, Claude Opus 4.8, and Gemini 3.1 Pro, found that combining multiple AI models does not create the safety net most enterprises assume. On the MATH-500 benchmark, standard correlation metrics predicted a 2.3% “co-failure rate,” the share of prompts where every model in a […]

Gen AI Is Making Compliance Teams Nervous: Here Is What the Data Shows

The Gap Between Adoption and Control A BCG analysis finds that as Generative AI adoption spreads across business units, 70% of legal, compliance, and privacy leaders now rank rapid GenAI adoption as their top operational concern for the next two years. Yet the same research shows firms integrating GenAI into administrative and compliance workflows have […]