{{first_name | Hi}}, your strategic AI update is here.
Expedia's AI chief said something today I fully agree with: models are basically swappable, the data is what's yours. Lenders are getting pickier - CoreWeave's bonds now yield 15% - and Replit's CEO says agents do a lot of his company's work. I'd stop adding AI tools this budget cycle and build on the stack you have, keeping every piece replaceable. Companies without a settled approach see strong AI returns 9% of the time, well behind everyone else.
Innovation Snapshot
"Sometimes you don't have to have the fastest car."
— AJ Wijesinghe, Chief AI Officer, Ecolab
In today's lineup
Robots
Shiny Object Syndrome is costing you AI returns
Real ROI from Gen AI
One list of every AI tool you pay for
People
Veeraj Shah takes hospital staffing off paper and Excel
Love
Give every business unit the same AI rulebook
Good Strategy Bad Strategy by Richard Rumelt
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
Expedia's AI chief calls models swappable, data the edge, keeping 80% of his people on one "golden path."
CoreWeave's bond rout tests the math on AI buildouts, as yields hit 15% and lenders get selective.
Replit's CEO calls his agent-run firm a "self-driving company", with human judgment kept in the middle.
IMF chief says AI risks may exceed its opportunities, without safeguards and timely productivity gains.
G7 economies plan simulated AI attacks on banks, a drill worth running on your own AI first.
Shiny Object Syndrome is costing you AI returns
Companies still trying out AI tools with no settled approach report the weakest returns in the new Wharton-GBK AI Adoption Report, a survey of 1,400+ people at large US companies. Just 9% of them see strongly positive ROI. Where the company standardizes on one approved platform, 50% do.
ROI leaders still test other tools, co-author Jeremy Korst told CIO Dive, but "the data seem to indicate that leaders have chosen a primary partner." With that settled, teams can spend their time changing how the work gets done.
Three moves for this budget cycle:
Settle how your company picks AI tools, and where exceptions are allowed
Build your next use case on the stack you already pay for
Keep data and workflows portable, so a model switch later doesn't mean a rebuild
The leadership pulse
How many AI platforms will you back in 2027?
Real ROI from Gen AI
How human-centric AI delivers business results
Southwest Airlines runs a Salesforce Agentforce assistant on its approved help content, resolving 45% of cases across 2M+ interactions, and sends safety and legal questions straight to a person.
Apollo Tyres lets an Amazon Bedrock agent set tire extrusion speed after each recipe change, with operators approving its adjustments at first, and cut startup rework by about 26%.
At Westpac, five AI agents built on AWS read payslips and apply lending policy, saving more than 150,000 hours, while bankers keep oversight and help customers decide.
Tool Spotlight
Your weekly briefing on tools that create competitive leverage
One list of every AI tool you pay for
Zylo keeps one live record of the software and AI you pay for. It recently made its Clarity AI agent available to all customers to find unused licenses on request, and its MCP server opens that record to AI tools like Claude and Gemini. Best for CIOs who want their own agents working from it.
Torii finds every SaaS and AI app in use and tracks AI spend by user and model. Freshworks agreed to buy Torii this week to add it to its IT service desk, so ask how Torii will work on its own after the deal closes. Best for CFOs who need AI costs charged back to teams.
Pick by who owns the problem
Zluri suits CIOs who need app discovery and identity governance for people and AI agents.
Nudge Security suits security teams who need to see AI apps employees adopt without going through IT.
Larridin suits Chief AI Officers who need to know which AI tools teams actually use before buying more.
PEOPLE 👥
Meet the innovators turning bold ideas into real-world impact.
Transformation Champion
Veeraj Shah takes hospital staffing off paper and Excel
Many hospitals still run their operations on paper and Excel. Veeraj Shah co-founded Vitalize, whose AI platform runs staffing and scheduling on top of a hospital's medical record and time-and-payroll systems, and is meant to replace more than 10 separate tools. Vitalize raised a $31M Series A earlier this month. At St. Luke's Health, premium labor spend on overtime and agency staff fell 54% in 12 weeks.
LOVE ❤️
Practical wisdom, growth tactics, and a must-read book that will challenge the way you think.
Give every business unit the same AI rulebook
Most organizations still struggle to agree on one AI standard across business units, FICO's chief analytics officer Scott Zoldi told CIO Dive. With a shared standard in place, he says, governance teams no longer stop deployments for not being sound or monitored. He calls that "the flywheel."
List the algorithms every business unit may build with.
Agree how each AI system gets explained and monitored.
Write down how each one meets regulatory requirements.
Transformative Reads
One book, handpicked from my conversations with friends, industry leaders, and tech innovators:
In Good Strategy Bad Strategy, Richard Rumelt argues that many strategies are really lists of goals. A good one names the main challenge, picks one way to deal with it, and lines up actions that support each other. Read it next to your 2027 AI budget and check whether the budget names a challenge or only lists tools.
Perfect for: CEOs and CIOs turning a pile of AI pilots into one plan.
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
