A previous article traced what IKEA, Nvidia, Klarna, and Duolingo built after AI returned capacity to their people. The pattern that emerged was specific: companies that decided in advance what to do with recovered time got materially different outcomes than companies that let the slack disappear into the existing workload. Now that same pattern has shown up in financial services, healthcare, consumer technology, and energy. Different industries, different use cases, different scales – and the same decision sitting at the center of each story.
Key Takeaways
- Morgan Stanley gave 20,000 financial advisors AI tools to handle note-taking and research synthesis – and the firm closed 100,000 new client relationships while growing total assets to a record $6 trillion.
- Kaiser Permanente recovered 16,000 hours of documentation time across 2.5 million patient encounters and is now building AI-assisted care navigation and chronic condition management programs that didn’t exist before.
- Airbnb automated 33% of US and Canada customer support, then used the freed engineering capacity to rebuild its entire platform – with AI now writing 60% of new code through Project Y.
- Spotify absorbed playlist curation and content recommendation at scale, then stood up two new businesses: Creative Lab, an in-house ad creative agency, and Quick Audio, a generative AI ad tool for advertisers.
- Octopus Energy had AI handling nearly half of all customer contacts by mid-2023 – and moved staff into energy advisory roles building an entirely new product category around smart energy management.
The Companies Getting the Most from AI Decided What to Do With the Time Before They Got It Back
Morgan Stanley’s deployment is the most documented financial services example. The firm gave roughly 20,000 financial advisors AI tools – built on GPT-4 – that surface over 100,000 internal research documents conversationally and handle post-meeting note-taking and follow-up task generation automatically. Advisors stopped spending time on document retrieval and administrative wrap-up. They started spending that time on clients. The results: 100,000 new client relationships closed, and total client assets grew to a record $6 trillion (Morgan Stanley, 2024 Wealth Management Report).
The growth didn’t flow from AI in a straight line – growth companies grow. But Morgan Stanley made a deliberate choice about where recovered advisor hours would go: toward new client acquisition and relationship depth. That choice was the frame the deployment was built around, not a conclusion reached after the tools went live.
Octopus Energy ran the same play in a completely different industry. By mid-2023, AI through their internal Kraken platform was managing roughly half of all direct customer support communications – transaction-level contacts that had previously consumed their customer service team. Rather than reduce headcount, the company moved staff into energy advisory roles: helping customers navigate EV tariffs, heat pump installations, and time-of-use energy plans. Those conversations require contextual judgment. The AI couldn’t have them. The people who used to answer routine billing questions now build the smart energy management products – EV charging optimization, heat pump scheduling, grid forecasting – that have become an entirely new product category for the business.
“Companies that decided in advance what to do with recovered time got materially different outcomes than companies that let the slack disappear into the existing workload.”
Redeployment Requires a Deliberate Transition, Not Just Available Capacity
Kaiser Permanente’s story illustrates the healthcare version of this. AI scribes – deployed across 40 hospitals and 600+ medical offices through partnerships with ambient documentation platforms – now handle clinical note-taking in real time during patient encounters. That freed 16,000 hours of documentation time across 2.5 million patient encounters (Kaiser Permanente, 2023 Innovation Report). Physicians got presence back: more eye contact, more listening, more of what drew them to medicine in the first place.
Kaiser is now expanding into AI-assisted care navigation and chronic condition management programs that didn’t exist before the documentation burden lifted. But that expansion didn’t happen automatically when the hours appeared. The transition required stability first – time for the new documentation workflow to settle before redirecting capacity into genuinely new work – and in some cases, additional training for the people moving into those expanded roles.
That sequencing matters more than most companies expect. People moving from high-volume transactional work into judgment-intensive advisory roles need a different skill set: critical thinking, problem-solving, systems-level perspective. The transition isn’t clean, and some of it breaks in the middle. But redeploying internal people rather than hiring externally carries a real structural advantage – they already know how the organization works, they carry institutional knowledge about the processes that preceded the automation, and that knowledge becomes critical when the new systems need adjustment.
New Business Lines Are Built on the Capacity AI Returns – at Every Scale
Airbnb’s case involves two separate AI deployments serving the same thesis. AI now handles 33% of US and Canada customer support – refunds, booking disputes, itinerary changes – without human involvement. Separately, AI writes roughly 60% of Airbnb’s new code. Engineers used the development capacity freed by AI-assisted coding to attempt Project Y, a ground-up platform redesign that CEO Brian Chesky has described as the most ambitious rebuild in the company’s history. That project would not have been feasible at that pace without it.
For companies watching this, the sequencing question matters: which deployment comes first? Start with engineering. When AI is writing most of the code, every downstream project moves faster. The team of engineers using AI to build becomes the resource that relieves bottlenecks across the entire organization.
Spotify’s move is structurally similar but creates new revenue rather than new product capability. AI absorbed playlist curation, content recommendation, and ad production at scale. Spotify pointed human teams at two new businesses: Creative Lab, an in-house creative agency producing branded audio content and campaign work for advertisers, and Quick Audio, a generative AI ad tool that lets small and mid-sized advertisers create finished audio ads without recording studios or voice talent. Both businesses serve customers who couldn’t access Spotify advertising before. Neither existed before the automation.
Mid-market companies can execute a version of this – probably one new line rather than two, and at a smaller scale. The economics work because automation typically generates revenue growth that funds the new venture. Companies can wait for that revenue to accumulate, borrow against it, or retrain people while the new line builds.
The Pattern Across Five Industries
| Company | AI Absorbed | Capacity Redirected To | New Outcome |
|---|---|---|---|
| Morgan Stanley | Note-taking, document retrieval, research synthesis | Client acquisition and relationship deepening | 100,000 new clients; $6 trillion in assets |
| Kaiser Permanente | Clinical documentation across 2.5M patient encounters | Patient presence; care navigation and chronic condition programs | 16,000 hours recovered; new care programs launched |
| Airbnb | 33% of US/Canada customer support; routine code generation | Full platform rebuild (Project Y) | AI writes 60% of new code; platform redesign underway |
| Spotify | Playlist curation, content recommendations, ad production | Two new business units | Creative Lab and Quick Audio both launched |
| Octopus Energy | ~50% of all direct customer contacts | Energy advisory roles; smart product development | New product category: EV optimization, heat pump scheduling, grid forecasting |
Frequently Asked Questions
Do these examples only apply to large enterprises, or can a mid-market company actually follow this pattern?
The scale differs, but the structure holds across company sizes. A mid-market company typically builds one new service line rather than two, and the initial deployment covers one team solving one clearly defined problem rather than 20,000 users simultaneously. The sequencing is the same: stabilize the automation, identify where recovered capacity goes, retrain the people who will do the new work. The funding model works at smaller scale too — automation generates revenue growth, and that growth funds the next move. [Link: AI implementation roadmap for mid-market companies]
What actually has to be true before a company can redirect recovered capacity into something new rather than just absorbing it into existing overhead?
Two conditions need to be in place. First, the new automation needs to be stable — roughly a month of settled operations before redirecting attention. Second, the people moving into new roles often need additional training. Redeploying internal staff over external hires carries a structural advantage: those employees already carry institutional knowledge about how the organization works and how the old process ran. That knowledge becomes critical when the new system needs adjustment or the transition hits friction. [Link: AI workforce transition planning]
Does the AI have to be fully deployed internally before a company builds new business lines, or can those happen simultaneously?
Sequencing matters here. The internal deployment needs to be stable and producing consistent results before the new work gets serious resources. Running both tracks simultaneously creates a situation where neither is properly defined, success criteria stay unclear, and recovered capacity never consolidates enough to build from. The sequence that works: solve one problem completely, define what success looks like, confirm you’ve hit it, then move.
About the Author: Audrey Kerchner is an AI strategy advisor at Inkyma who helps mid-market companies move from AI experimentation to AI execution – building the systems that return real capacity to their people and identifying where that capacity can go next. She works directly with leadership teams on deployment sequencing, workforce transition planning, and building the new business lines that recovered capacity makes possible.
The pattern in this article plays out at every scale, including yours. Schedule A Strategy Session to talk through where your business actually has room to grow.
