Blog

round mirror with reflection of man standing on the ground

Author and Columbia professor Halle Tecco recently shared a thought provoking <a data-mce-href="https://www.linkedin.com/posts/halletecco_i-recently-spoke-with-a-hospital-cmio-who-activity-7453043817009676288-1O3Z?utm_source=share&amp;utm_medium=member_desktop&amp;rcm=ACoAAAADleoBbUgcAt8rZFR-egBkNo3efiATkhg" href="https://www.linkedin.com/posts/halletecco_i-recently-spoke-with-a-hospital-cmio-who-activity-7453043817009676288-1O3Z?utm_source=share&amp;utm_medium=member_desktop&amp;rcm=ACoAAAADleoBbUgcAt8rZFR-egBkNo3efiATkhg" target="_blank" rel="noopener">post</a> on the power that Epic exerts on the healthcare industry, and on healthcare startups and innovation in particular.&nbsp; Here's the text of the post:"I recently spoke with a hospital CMIO who told me they were switching AI vendors. They were moving from a well-funded AI scribe startup to Epic's native AI ambient scribe, even though he genuinely liked the startup's product better.&nbsp;It came down to money: Epic was $100K a year compared to the startup’s annual cost of $1M.That story is playing out across the industry. SaaS is struggling broadly; software stocks lost over $1 trillion in early 2026, in part because it's easier to build things in-house now.&nbsp;But in healthcare, we also have the disadvantage of having an incumbent like Epic that squeezes out competitors. In Robert Wachter's words: "Epic’s stranglehold on the healthcare system AI market is likely to stifle investment and innovation".&nbsp;I do think the “SaaSpocalypse,” as some people call it, is an overcorrection. But if your entire pitch is a workflow automation layer on top of someone else's data, you're in a tough spot."This is spot on,&nbsp; but while I'm skeptical about the SaaSpocolypse, as I've written before, I would say it's more dire for SaaS &amp; AI startups in healthcare vs. any other vertical because of the stranglehold Epic has on the industry.Healthcare by its nature (necessity?) is massively risk averse, and the explosion in ransomware attacks in the late teens &amp; early 20's have only hardened that mentality relative to technology adoption.To echo what the CMIO told Halle, I've personally heard the following from CIO's, CISO's and CMO's for both regional systems and large IDN's (paraphrasing from memory):<ul><li>- "If I know something is on Epic's roadmap, I'm either not buying from another vendor, or will only sign short term deals so I can switch when Epic is ready"</li><li>"I want to buy platforms, I'm not interested in point products anymore"</li><li>"My CEO won't question my decision to buy from Epic/Oracle" (twist on the old-school Silicon Valley truism that "nobody ever got fired for buying Cisco...")</li></ul>Over $1.5 billion-with-a-B of capital has gone into ambient listening startups in the last couple years, and 2 months ago Epic released AI Charting. I wonder what the return prospects are now for those investors?The advice I'm giving to founders is if you think Epic can build a competitor to you, pivot. Doesn't matter how much better you believe your tech is, healthcare doesn't buy cutting edge or disruption. Plus, you will always be at a price disadvantage. Instead, focus on how you can build a moat that Epic can't or would choose not to try to bridge.&nbsp;In many (most?) cases that's easier said than done, but to build sustainable adoption, that's the only choice unless &amp; until Epic's stranglehold loosens.

The Epic Threat to Innovation in Healthcare

I just read Pete Flint's piece, <a data-mce-href="https://www.nfx.com/post/infinite-game" href="https://www.nfx.com/post/infinite-game" target="_blank" rel="noopener">The Infinite Game: What Artists Know That Founders Don’t,</a> and it got me thinking...again.&nbsp;Pete's thesis: Software is having its daguerreotype moment (yes, I had to look that up too) thanks to AI. When photography first appeared in 1839, painters thought it was the end. Who needs painted portraits when a machine can capture IRL perfectly? But in reality, photography freed artists from spending time on literal reproduction/documentation and instead allowed them to focus on creative expression...the 'art' of painting. At the same time, photography itself exploded through photojournalism, artisitic photography, instant Polaroids. The world of visual expression didn't contract. It expanded in directions no one predicted, creating new professions, markets, and industry segments.<figure id="figure-wy5axfvqpmonjqoak"data-full-width><img id="wy5axfvqpmonjqoak" src="https://cdn.durable.co/covers/30SoRpiPpkaRMGC3SRHV4bEy9rLTyo7S9y43lUkekEb0qvnyD2mHEuaoRerrEX3b.jpg" alt="" title="" class="object-cover bg-gray-100 rounded-sm md:rounded-md lg:rounded-lg" style="aspect-ratio: 16/9; object-position: center center;" data-corners="default" data-aspect="16:9" data-caption="" data-positionX="undefined" data-positionY="undefined" data-fullWidth="true"/></figure>I believe we are at an identical inflection point with AI. The doomsayers focus on what's being commoditized - basic coding, rote tasks, operational patterns - but that's not destruction, it's liberation. AI handles the execution layer, which frees up cycles for humans to exercise creativity and tackle more opportunities.&nbsp;Interestingly, the speed of adoption in healthcare has so far been mind-boggling given the industry's historical aversion to deploying new innovation. IMHO, that's mainly due to existential threats to the industry that threaten its financial viability (posts for another day) and ability to provide the services our society needs.Take AI scribes. They've very quickly gone from 'what's ambient listening?' to standard-of-care level tools because the data very quickly showed them to be effective in countering one the most significant existential threats - provider burnout. A study published in JAMA Oct-25 titled <a data-mce-href="https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2839542" href="https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2839542" target="_blank" rel="noopener">Use of Ambient AI Scribes to Reduce Administrative Burden and Professional Burnout</a> found that after just 30 days of use, providers experiencing burnout dropped from 51.9% to 38.8% because AI scribes drive meaningful drops in after-hours charting and cognitive load, and ultimately free clinicians to focus on patients instead of paperwork. I have even heard from C-level execs from the largest hospital systems that even though AI scribes are a net-loss, they invest in them anyway because it's the only way to recruit and RETAIN the best physicians. Significant statement in an industry with low single digit net margins.And it's not just about provider admin automation. There is massive innovation happening in diagnostics, drug development, personalized medicine, and all over the back-office.&nbsp;I'll go out on a limb - 10 years from now, there will be more people in healthcare than there are today, but they will be focused on clinical tasks that actually improve patient outcomes, instead of EMR charting, compliance, coding or other 'AI-able' admin tasks. In retrospect, EMRs will be the daguerreotype of the healthcare stack.Read Pete's piece and come back &amp; tell me what you think.

Software's Daguerrotype Moment?

The letters ai are displayed on a blurred background.

Harken back to a post I wrote a few months ago titled Is SaaS Eating AI?, Tomasz Tunguz nails it again in his <a data-mce-href="https://www.linkedin.com/posts/tomasztunguz_in-1908-253-american-automobile-manufacturers-ugcPost-7421658103433211904-O1cp/" href="https://www.linkedin.com/posts/tomasztunguz_in-1908-253-american-automobile-manufacturers-ugcPost-7421658103433211904-O1cp/" target="_blank" rel="noopener">analysis</a> of the likely impact of AI on jobs - especially tech sector jobs.TL;DR the sky isn't falling (I'm pretty sure...).Here's the crux of the analysis - In 1909 there were 253 auto manufacturers. Then Henry Ford introduced the assembly line, and 20 years later only 44 manufacturers remained. But jobs in the industry over that same period went from 76K to 471K.&nbsp; That's over 6X (??) increase of jobs in the sector.&nbsp;<figure id="figure-94vdvrhu4monjbqok" data-full-width=""><img id="94vdvrhu4monjbqok" src="https://cdn.durable.co/blocks/27LsYCo7dZgASOjUMueXRo99NTnNuPeuTA3LMRskdjR6qusQ5UPNdyWMZI3UeoQe.jpg" alt="" title="" class="object-cover bg-gray-100 rounded-sm md:rounded-md lg:rounded-lg" style="aspect-ratio: 3/2; object-position: undefined undefined;" data-corners="default" data-aspect="3:2" data-caption="" data-positionx="undefined" data-positiony="undefined" data-fullwidth="true" data-mce-src="https://cdn.durable.co/blocks/27LsYCo7dZgASOjUMueXRo99NTnNuPeuTA3LMRskdjR6qusQ5UPNdyWMZI3UeoQe.jpg" data-mce-style="aspect-ratio: 3/2; object-position: undefined undefined;"></figure>What happened? Automation fundamentally changed throughput, supply chain and unit economics. The industry not only shifted to feed the new beast, but jobs in it grew by 6X+ over the same time.&nbsp; Jobs didn't go away - they shape shifted &amp; multiplied.Feels familiar. Secondary &amp; tertiary effects of industrialized AI are just barely beginning to emerge. But they will. And they'll likely follow the same pattern that followed the assembly line.&nbsp; As factory automation drastically reduced both the production time and cost of cars, AI will do the same for technology (coding, devops, etc.).&nbsp; But the resources that would have otherwise been operating in pre-AI type roles will redeploy to&nbsp; build and build the ecosystems &amp; supply chains necessary to support the post-AI dynamics of building and operating technology.<figure id="figure-wlxl94ktc9monjgyfr"><img id="wlxl94ktc9monjgyfr" src="https://cdn.durable.co/blocks/26w8OUkbMKlBnvVkhf1nZAK3rwEjN7MgOKYvRQSLTAPyr6Q2jfq9VqNUAvMpDqS3.jpg" alt="" title="" class="object-cover bg-gray-100 rounded-sm md:rounded-md lg:rounded-lg" style="aspect-ratio: 3/2; object-position: center center;" data-corners="default" data-aspect="3:2" data-caption="" data-positionx="null" data-positiony="null" data-fullwidth="false" data-mce-src="https://cdn.durable.co/blocks/26w8OUkbMKlBnvVkhf1nZAK3rwEjN7MgOKYvRQSLTAPyr6Q2jfq9VqNUAvMpDqS3.jpg" data-mce-style="aspect-ratio: 3/2; object-position: center center;"></figure>I'm old enough to remember the cries that the cloud was going to destroy the enterprise hardware business. Nvidia would like a word.&nbsp;From Tomasz' original post here - <a data-mce-href="https://www.linkedin.com/posts/tomasztunguz_in-1908-253-american-automobile-manufacturers-ugcPost-7421658103433211904-O1cp/" href="https://www.linkedin.com/posts/tomasztunguz_in-1908-253-american-automobile-manufacturers-ugcPost-7421658103433211904-O1cp/" data-mce-selected="inline-boundary">https://www.linkedin.com/posts/tomasztunguz_in-1908-253-american-automobile-manufacturers-ugcPost-7421658103433211904-O1cp/</a>

Is AI Kiling Jobs?

smiling girl in black and white striped shirt

To my fellow entrepreneurs, CEO's, marketing and Cx leaders, I'm sharing this as a cautionary tale of 'how not to' provide outstanding service. But I'd guess most of you already know &amp; feel this.An adversarial customer service experience is one of the surest-proof ways to destroy your brand along with any possibility of ever winning back a lost customer in the future. Take it to the bank and build accordingly.Here's the set-up. I've been a Cox Communications broadband customer since I moved to AZ. 11.5 years and easily over $20K spent with them over that period. They don't offer fiber in my neighborhood, and have increased rates by almost 65% in the last 5 years for the same speeds/service level for HFC. So did some digging and turns out not only is there a fiber provider in my neighborhood, my house is already prewired for it.3X the speed and it costs 60% less than what I'm paying now. No brainer.Install was today. Seamless, 45 minutes and the tech was smart, polite and my dogs liked him. Bonus.Spend 5 minutes searching cox.com for how to cancel. Finally find it, and click 'chat with us'. Here's how it should have gone, in 5 minutes or less:Cox: It's great to connect with you today Andrew. How can I help you? Me: Thank you. I'd like to cancel my service. I've switched to a different ISP. Cox: I'm terribly sorry to hear that. I see you've been a loyal customer for over 11 years. Anything I can do to change your mind? Me: No, the service has had spotty reliability, I'm looking for higher speeds and you don't offer fiber in my area. Cox: Understood. Let me process your request. ..... Cox: OK, the cancellation is complete [insert confirmation info]. Would be OK for us to reach out to you once we do have fiber in your area to see if we can win you back? Me: Absolutely. Thanks for your help. Cox: My pleasure. Thank you again for being a loyal customer for so long.Brilliant. You can't offer me what I want at the moment, but the positive exit experience leaves the door open for you to win me back when you can.Seems so obvious. But, what followed instead was an excruciating 57 minutes of argumentative offer hounding (clearly with google translate in the middle based on grammar &amp; response times) that p*ssed me off and felt more like I was negotiating with a street hawker in Marrakesh to buy a fake Rolex vs. dealing with a multi-billion dollar communications company.In a world of Cox Communications, be Chick-fil-A

The Perils of Bad Customer Service

Here's a link to another fantastic <a data-mce-href="https://www.linkedin.com/pulse/great-startup-delusion-what-web3-failures-teach-business-paul-o-brien-lu5ac/" href="https://www.linkedin.com/pulse/great-startup-delusion-what-web3-failures-teach-business-paul-o-brien-lu5ac/" target="_blank" rel="noopener">post</a> from my friend Paul O'Brien.&nbsp; This time he's discussing something that seems pretty straightforward, but isn't - what is a startup?There's a tendency for anyone building anything with technology to claim they're building a 'startup,' but it's not that simple.&nbsp; Here's the framework from Paul's article that sticks out to me:"Startups are not just new businesses. Startups are new business models.They’re not “another version” of an existing idea with tech sprinkled on top. A startup creates an entirely new kind of economic behavior, often enabled by innovation but not reducible to it. It changes how markets operate, how customers behave, or how industries are structured. It breaks something, reconfigures it, and creates disproportionate value by doing so."I've posted my thoughts on startup ecosystems before, and I believe this lack of clarity lies directly at the heart of a lot of the startup ecosystem failures we talk about.&nbsp; Every startup is a small business, but few small businesses are really &amp; truly a startup.Incidentally, this reminds me of my favorite definition ever given for a 'startup,' which is (to paraphrase the great Steve Blank):"A startup is a temporary organization in search of a repeatable business model."Note the complete absence of the word 'technology'.&nbsp;Thanks as always Paul, incredibly insightful.Go read Paul's post &amp; then let me know, how do you define a startup?

What is a Startup Anyway?

I just read through this MIT report on "The GenAI Divide: State of AI in Business 2025" (link in comments). While I'm not entirely surprised at the data, it's still an eye opener and a good reminder that while we're knee (neck?) deep in an AI hype cycle, it's far from delivering on its promises—for most organizations, anyway.&nbsp;<h3>TL;DR - Be patient.</h3>The study, based on interviews with 52 orgs, surveys from 153 leaders, and analysis of 300+ AI initiatives, paints a picture of massive investment ($30-40B in enterprise GenAI) but shockingly low returns: 95% of companies are seeing zero measurable P&amp;L impact. Tools like ChatGPT are everywhere—80% explored, 40% deployed—but they're mostly boosting individual productivity, not transforming businesses.&nbsp;The money quote - "Enterprise grade systems, custom or vendor-sold, are being quietly rejected. Sixty percent of organizations evaluated such tools, but only 20 percent reached pilot stage and just 5 percent reached production. Most fail due to brittle workflows, lack of contextual learning, and misalignment with day-to-day operations." The outcome from $40B invested (so far) - stalled pilots and a "shadow AI economy" where employees sneak in personal tools for real gains but enterprise-wide implementation remains elusive.Not surprisingly, the media industry is moving the fastest toward broad adoption given the incredible pace of improvement in GenAI media tools, but most of the rest of the economy is so far showing minimal disruption. Budgets skew mainly toward sales/marketing automation(50%+), while back-office and ops — where ROI could be huge — remain largely overlooked.&nbsp;Reminds me of the early days of cloud computing. I'm old enough to have lived through &amp; remember the hype when AWS launched in 2006 but concerns over security, data sovereignty, and integration meant a slow evolution from on-prem to the cloud. It took about a decade for cloud to hit critical mass: by 2015-2016, market share started tipping (AWS alone crossed $10B revenue).&nbsp;The lesson? Disruptive tech needs time to mature - refining models, building trust, and proving ROI - before it reshapes markets.GenAI feels similar: we're in that frothy, experimental phase, but the report highlights paths forward for "crossing the divide." So, while the hype is real and the returns aren't (yet), like cloud, GenAI will redefine efficiency and innovation once we get the fundamentals right.What do you think? Have any war stories or cautionary tales to share? Let me know.

Insights

The Epic Threat to Innovation in Healthcare

Software's Daguerrotype Moment?

Is AI Kiling Jobs?

The Perils of Bad Customer Service

What is a Startup Anyway?

Crossing the AI Hype Chasm