Insights

What we think, in writing.

Research, white papers and opinion on putting AI into production and measuring what it returns.

Dispatches

New research, when it publishes.

Occasional dispatches on measurable AI.

Unsubscribe any time. Privacy

Manifesto4 min readPeter Grant, CEO, Weaver

The bus stops here.

The Big 4 built their AI practices the same way they built everything else, by pulling the bus up and filling it with bodies. There is a better way.

Read the manifesto
  1. 20White paper · 14 minOwn the answer: A decision framework for closed, open, cloud, and on-prem AI.The question has moved from which model is smartest to who owns the weights, where they run, and who controls the data that shapes them.
  2. 19White paper · 10 minAn introduction to decoupled agent frameworks.At organizational scale, the unit of work is an agent, rarely a single prompt. Scaling those agents rests on an old engineering principle, decoupling, applied to AI.
  3. 18White paper · 12 minFour architecture decisions that get enterprise AI to production.95% of business AI programs fail to deliver measurable business value. The models are rarely the bottleneck. The architecture around them is.95%AI programs with no measurable valueSource · MIT NANDA, 2025
  4. 17Research · 9 minAI literacy is the multiplier. Role-based training is how you earn it.Ninety-five percent of business AI programs fail to deliver measurable value. The cause is usually the workforce, not the model. Role-based enablement closes that gap. Generic training does not.
  5. 16Opinion · 7 minWhy 40% of agentic AI projects will fail by 2027. And what the other 60% are doing differently.Agentic AI failure is an architecture problem, not a talent or strategy problem. Deloitte's own data says so, and the fix isn't more consultants.
  6. 15Opinion · 6 minYour AI partner just became a Google salesperson. Here's what that means for your architecture.When your advisor's compensation depends on what stack you choose, you no longer have an advisor. You have a salesperson with a McKinsey badge.
  7. 14Opinion · 6 minThe 80/20 rule that PwC won't tell you about, and why it argues against hiring them.PwC's own research says technology delivers 20% of an AI initiative's value. The other 80% is organizational. So why are enterprises paying Big 4 day rates for a technology deployment?
  8. 13Opinion · 7 minThe 40% problem: What MIT and BCG discovered about AI and the workers using it.When AI is used inside its capabilities, knowledge worker performance jumps by 40%. When it is used just outside them, performance drops by up to 24 points. The line between the two is invisible. Finding it is the job.40%Performance gain inside the AI frontierSource · MIT, Harvard and BCG
  9. 12Opinion · 7 minYou're probably using 10x more AI than you need to.MIT's Lottery Ticket Hypothesis showed that inside every large neural network, a tiny subnetwork is doing almost all of the work. The same principle is sitting inside your AI invoice, and most enterprises are still paying for the whole ticket book.
  10. 11Research · 6 minThe adoption gap: Why trained users get 15x more from your AI investment.Without training, the most sophisticated enterprise AI platform becomes an expensive search engine. With it, the same platform delivers 3.2x the engagement and 15x the use of advanced capabilities.
  11. 10Opinion · 5 minThe silicon divide: Why your corporate laptop is strangling your AI ROI.If agentic AI tools save a knowledge worker 30 to 60 minutes daily, the $1-a-day hardware premium pays back in the first week. Most IT departments have not run this number.
  12. 09Case study · 5 minHow The Telegraph and Mimecast put AI into production, and kept it there.A 64 to 94% reduction in first-pass editorial edit time. A 20% increase in sales enablement content productivity. Production numbers, not pilot results.
  13. 08Opinion · 4 minYour AI vendor will deprecate their model. Here is why that should not be your problem.The AI landscape shifts every 90 days. Most enterprises have no protection against what happens when it does.
  14. 07Research · 6 minYour AI pilot worked. You just can't prove it.Most AI pilots succeed technically and fail commercially, because nobody defined what delivery meant before they started.
  15. 06Opinion · 8 minRAG is not dead. And if you believe otherwise, you are about to make an expensive mistake.The "RAG is dead" story appeals because it promises simplicity: one model, one giant context window, no messy pipelines. For small, static datasets, it can even be true. Enterprises are not small or static.
  16. 05Opinion · 3 minThe $186 hidden tax your AI program is generating every month.Your organization is paying $186 per employee per month to fix AI's mistakes. Most CxOs have no idea this number exists.
  17. 04Opinion · 3 minWhy McKinsey moving to performance fees validates everything we said.McKinsey is shifting to performance-based fees because AI is making the billable-hour model indefensible. The industry knew this was coming.
  18. 03Research · 5 minAI returns $3.70 per dollar. Most organizations get less than $1.The gap between AI leaders and laggards is architecture and governance, not budget or model selection.
  19. 02Research · 6 minWhy most enterprise AI dies between the demo and production.The demo works. Then it meets real data, real systems and real users, and 95% do not survive.95%Enterprise AI pilots that never reach productionSource · MIT, 2025
  20. 01Research · 4 minMcKinsey deployed 12,000 AI agents. Here is what enterprise AI actually looks like at scale.When McKinsey reduced one consulting team from 14 people to 2 to 3, the industry could no longer pretend this was not happening.
Appearances

In conversation.

Weaver's founders as guests on other people's shows.

  • Aug 2026BBC NewsWhy isn't AI working for your company?BBC News examines why so many enterprise AI programs fail to deliver measurable returns, and what separates the organizations getting value from the ones still stuck in pilots.Watch ↗
  • Apr 2026TECHtonicDoug Duker on TECHtonic with Thomas LahThomas Lah (Executive Director of TSIA) and Doug Duker, then Global Leader of Customer Success at You.com, on what sets AI-native companies apart from traditional SaaS: data security, multi-model agility, hallucination control and personalized AI training.Listen ↗
  • Apr 2026ToplinePeter Grant on the Topline podcastPeter Grant on the operational reality of running an AI-first go-to-market organization, and why the current sales playbook is already obsolete.Listen ↗
  • Apr 2026GTMnowPeter Grant on the GTMnow podcastPeter Grant, then CRO of You.com, on hiring elite go-to-market talent and how to make AI literacy operational to close the ROI gap.Listen ↗
  • Apr 2026Make It Happen MondaysPeter Grant with John BarrowsHow AI is reshaping enterprise sales, why AI literacy is becoming a required skill, and why the gap between top performers and average sellers is about to widen.Listen ↗
  • Apr 2026Revenue InsightsPeter Grant with Guy RubinWhat it takes to scale in an AI-first world: the four signals that separate real AI companies from hype, and the shift from information retrieval to decision enablement.Listen ↗
The next dollar

Know which AI investments deserve the next dollar.

Deliver your Return On Intelligence.