The AI ROI gap is a forecasting problem

THE AI BRIEF

Today's signal: Bain says the AI deployments worked, but the cost savings didn't show up. Plus Alphabet's $80B raise, Coralogix's $200M, the Trump AI EO, and a Microsoft token-efficiency claim.

In today’s issue:

• Main story: Bain says the AI worked, but the savings did not show up

• Also worth knowing: Alphabet raised $80 billion in equity, Coralogix raised $200 million to monitor AI agents in production, the Trump administration creates a voluntary 30-day model preview program, and more.

THE READ

A new Bain survey reports that most enterprise AI deployments are functioning as intended and missing the cost-reduction targets that were used to justify them. The interesting part is what that does to the second wave of pilots.

A Bain & Company report covered by Bloomberg on June 1 summarized its findings in one line: "The technology worked. The value didn't arrive." The survey found that most enterprise AI deployments are running as designed, the models are doing what the implementations were supposed to do, and the cost-reduction numbers that boards approved are not landing on the income statement on the timeline that was promised. The story has been the highest-engagement enterprise AI item this week, picking up more than 450,000 views and a thousand likes across the social cuts that have been circulating since the weekend.

The framing matters. This is not the older story of a failed pilot or a model that hallucinated through a customer-service script. The systems work. The headcount savings, the cycle-time reductions, and the procurement consolidations that were modeled inside the business cases are not appearing where the finance team expected them. Bain's read is that the forecasts were too optimistic, not that the technology was. The gap is in the planning layer, not the engineering layer.

AI READY PRO · FREE UNTIL FRIDAY

I will get you AI trained in 30 days for free

After several months of running it quietly with top AI operators and teams, I’m excited to launch AI Ready Pro today to our newsletter subscribers like you.

It’s a 30-day personalized AI training program that is personalized to you and where you are in your AI journey. It’s the culmination of 100’s of hours spent teaching AI to top Fortune 500 AI teams and operators (and working with Mark Cuban).

Each day, you receive a 10-15 minute exercise to complete to improve your AI skills.

This isn’t vague AI theory or a way to pitch you a tool. It’s 30 days of learning to use AI in your real work. So you come out the other side with the skills to use AI to actually improve you and your teams output.

Until this Friday (May 15), it’s free for newsletter readers who complete the extended assessment. This assessment will help us personalize the learning experience to you.

If you lead a team, AI Ready Team is open today too, also for readers first. It’s the same engine, but to train your whole organization. You’ll get a detailed view of:

• Your team’s AI readiness level + detailed strengths & weaknesses

• Ranked list of AI automation opportunities

• Get tactical advice on how to advance AI efforts in your org

Leading a team? Take the Team assessment instead.

For a mid-market operator, the practical question is whether a working pilot that misses its savings target is treated as a pilot success or a pilot failure when the second-round funding decision comes up. The default reading inside most boards I have been in is that a deployment that runs and does not produce the modeled cost savings is the same as a deployment that did not run, which is the wrong reading. The right reading is that the cost-savings model was wrong and the deployment is now a base from which a different value case can be built, usually on revenue or quality rather than headcount. What I keep hearing in client conversations is that the teams that get the second round of budget are the ones who reframe the outcome before finance does, and the teams that lose the budget are the ones who let finance score the pilot against the original cost model.

The bear case is that the Bain finding is being read more dramatically than the report supports. A consulting firm telling enterprise buyers that AI is missing its cost targets is a finding with a built-in audience, and the survey methodology and sample size matter more than the headline does. The thing worth tracking over the next quarter is whether the public AI ROI story from named companies starts shifting from cost reduction to revenue and quality language. If it does, Bain called it early. If it does not, the survey is closer to a snapshot than a turn.

SPONSORED BY CLUTCH

Hire secure AI teammates that work 24/7.

Hire pre-built AI teammates. Give your engineers and operators a platform to ship their own AI apps. Stop losing sleep about what is running where.

Clutch is the platform behind both: pre-built agents for the workflows your ops team should automate first, plus the integration plane your team's vibe-coded apps and Claude Code projects plug into. One platform. Real production. Visible and safe by default.

Built for ops, engineering, and security teams that are tired of the shadow-AI surface area inside their own company.

ALSO WORTH KNOWING

Alphabet raised $80 billion in equity, and the stock fell 7%. Analysts noted that roughly half the raise is earmarked for employee stock tax obligations, leaving about $40 billion as fresh AI infrastructure spend. The S&P 500 hit a record in the same session.

Coralogix raised $200 million to monitor AI agents in production. The round funds observability tooling specifically for autonomous agents, where traditional monitoring stacks fall short of explaining decisions, reasoning chains, and failure modes. Agent monitoring is now a category with venture-scale capital behind it.

The Trump administration signed an executive order creating a voluntary 30-day model preview program. Frontier AI labs that opt in will share pre-release model access with federal agencies for evaluation 30 days before public launch. The opt-in design keeps the friction low for developers while giving agencies a head start.

SambaNova demonstrated a disaggregated inference architecture at Computex. NVIDIA B200 GPUs handle input processing, and SambaNova RDUs handle output generation, with the company reporting 2x throughput against a B200-only baseline. The signal is that the inference stack is splitting into specialized layers.

Microsoft added average token usage to the MAI-Code-1-Flash model card. The model hits 71.6 on SWE-Bench Verified using roughly a third of the tokens, Claude Haiku 4.5 burns. Tomasz Tunguz called the token disclosure "a new standard," and other labs are likely to follow.

Martin Scorsese joined Black Forest Labs as an advisor on FLUX. Scorsese is using the visual model for film pre-production storyboards and has been credited as a creative advisor on the next FLUX release. It is the highest-profile creative-industry endorsement a generative image lab has landed.

WATCHING TOMORROW

Microsoft Build wraps Thursday afternoon with the Agent 365 and Frontier Tuning sessions, which are the parts of the keynote stack that name enterprise customers. Friday brings the May jobs report, which is the first labor read of the post-Q1 AI capex cycle and the data point most likely to move the AI-and-employment conversation next week.

Back tomorrow,
Haroon