{{first_name | Hi}}, your strategic AI update is here.
American Express is cutting 5% of staff today - and its CFO put it well: don't automate processes that aren't well designed. Anthropic is backing a consultancy to embed engineers in European firms, and SpaceX is funding its AI build-out with debt. Everyone needs AI to start paying back. I'd give every use case a number to move and a date to stop it if it doesn't. Only 31% of CIOs have that rule for pilots, Gartner found.
Innovation Snapshot
"Winners quit all the time. They just quit the right stuff at the right time."
— Seth Godin, Author of The Dip
In today's lineup
Robots
Pilot Purgatory needs a written exit rule
Real ROI from Gen AI
Teach your agents what revenue means
People
Linda Du brings auditable AI agents to mortgage operations
Love
Choose what your AI reads at decision time
The Phoenix Project by Gene Kim, Kevin Behr and George Spafford
Reading time: 3 min.
ROBOTS 🤖
How are robotics and AI changing industries? We break down the latest news, tools, and innovations for you.
Top Insights
American Express cuts 5% of staff ahead of AI, with its CFO warning against automating processes that aren't well designed.
Anthropic and Hg set up a European AI consultancy, a sign CIOs need help deploying AI more than new tools.
SpaceX won't sell new shares to fund AI, JPMorgan says, more AI capacity built on debt, so keep your stack portable.
PG&E's CEO calls AI the best thing for the grid, as tech money pays for the power AI-heavy sites need.
The SEC will examine how Wall Street firms use AI, so keep a record of who supervises each AI system.
Pilot Purgatory needs a written exit rule
"Volume is not value," Gartner's Alicia Mullery told CIOs at Gartner IT Symposium/Xpo this week. Of the 2,002 CIOs surveyed, only 13% say AI has delivered significant value so far, and about one use case in five succeeds.
Most CIOs haven't decided in advance when to stop. Just 31% have guidelines for when a pilot should not go to production, and 17% for when to drop low-value AI that's already live. CIOs with these rules are 2.3 and 2.6 times more likely, respectively, to be in Gartner's top group.
Mullery added that AI costs, unlike cloud or SaaS, are decided at run time. A weak use case that's live keeps costing money until someone turns it off.
Before funding the next use case, agree with Finance on the number it has to move. If it hasn't moved after one twelve-week cycle, stop it and put the money into one that's working. IT and the business owner should make that call together.
The leadership pulse
What happens to an AI pilot that misses its number?
Real ROI from Gen AI
How human-centric AI delivers business results
Kodiak Gas Services technicians order parts through an agent on Microsoft Foundry, bringing each request down from 15-20 minutes to 2-3, and each order still needs approval.
Chatham Financial's trade-checking app, built with OpenAI Codex, cut review from about 30 minutes to under 4 in early tests, and reviewers still check its work.
Grupo Bafar's legal and retail teams prepare store-opening contracts with Gemini Enterprise, dropping prep time per branch from 140 to 17 minutes.
Tool Spotlight
Your weekly briefing on tools that create competitive leverage
Teach your agents what revenue means
AtScale gives every AI agent and dashboard one agreed definition of numbers like revenue and margin. Last week it launched AtScale for AI, which also tests AI answers against known results. HSBC's UK commercial bank reports AI analytics accuracy rising from 65% to over 99%. Best for Chief AI Officers who want agents to calculate revenue the way Finance does.
Honeydew builds a shared model of your business logic inside the warehouse you already run, such as Snowflake or Databricks, so Power BI, Tableau and AI agents all read the same definitions. It uses your own LLM provider and keeps a version history of every change. A good fit for CIOs who want one set of definitions without moving data to a new platform.
Three more semantic layers
dbt Semantic Layer keeps metrics as versioned code, a natural add-on for IT teams that use dbt.
Kyvos is built for very large data sets and governs metrics for AI and dashboards in one place.
Timbr gives agents business context from data spread across many platforms, for teams that can't consolidate first.
PEOPLE 👥
Meet the innovators turning bold ideas into real-world impact.
Transformation Champion
Linda Du brings auditable AI agents to mortgage operations
Before selling any software, Linda Du and her Valon co-founders spent six years running a mortgage servicer on their own platform. That platform now runs AI agents that answer homeowner emails, allocate payments and run escrow analyses, keeping an audit trail of everything they do. Two of the ten largest US servicers are live on it, and Valon closed a $150M Series D this month.
LOVE ❤️
Practical wisdom, growth tactics, and a must-read book that will challenge the way you think.
Choose what your AI reads at decision time
Think of your data lake as a warehouse. When an order lands, your AI needs a chef who grabs the right, fresh ingredients and nothing else. Point it at the whole warehouse, or at last month's copies, and a good use case can miss its number for the wrong reason.
List the exact data one AI decision needs, and where each piece lives.
Serve that data from systems you have today, with every piece replaceable.
Set how fresh each piece has to be, and flag it when it goes stale.
The warehouse-and-chef picture is from the context layer chapter of 100x.
Transformative Reads
One book, handpicked from my conversations with friends, industry leaders, and tech innovators:
Bill Palmer gets 90 days to rescue Phoenix, a late, over-budget IT project, or IT gets outsourced. Every emergency lands on one engineer, so nothing planned moves. In The Phoenix Project, a novel by Gene Kim, Kevin Behr and George Spafford, the fix starts with cutting work in progress. Read it if your team has more AI pilots open than it can close.
In Culture
THANK YOU!
Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.
We just ask for one thing: tell one friend about Lighthouse today!
Want daily AI & Innovation insights? Connect with me on LinkedIn or X.
Got suggestions? Just reply to this email!
Much Love,
Matt and the Future Works team
