Heidi is either a giant disruptive tech Kraken which is coming for every incumbent medical data and workflow and technology platform in the world, or it’s the cheeky monkey in Night at the Museum, and there’s a storm coming.
In March I travelled to Melbourne to, among other things, attend Heidi’s AI coming out party. I had requested an invitation, logged myself as from the media, and was accepted as a guest.
When I got through the long line to get in, I was pretty excited by what I was seeing inside: 600 or so people, mostly clinicians, all buzzing, a giant launch stage complete with a wraparound cinematic screen featuring various spectacular images of nature, hip hop music booming in the background, merch going off like it was a Britney Spears revival, great finger food (though no alcohol) and, most pertinently for me, about six to eight demonstration stations, all crowded around with clinicians, many mesmerised and asking good questions, of what was a pretty fascinating demonstration of Heidi’s workflow, including its relatively newly released decision support layer, Heidi Evidence.
There was a rave feel to the set-up and a very positive vibe. Not exactly what you’d expect of medical demonstration event. Digital Health Festival, eat your heart out.
At some point before the main stage show featuring newly minted rock star start-up CEO Dr Thomas Kelly, one of Heidi’s new recruits, who I knew from a previous role (of just a few weeks prior) let slip to me that there would be a big announcement on the main stage later.
I asked immediately what announcement, to which I was given a slightly puzzled look and a reply along of the lines of, “oh, I think it’s all a secret…”.
I explained myself. I’m from the media (which this person already knew), and I think I may have stuffed up by missing the embargoed press release and, possibly, the opportunity to organise myself for pre-announcement interviews.
Now, a new confused look from my Heidi friend. He said awkwardly that he would try talk to someone in comms and get back to me.
There was still about two hours before the big reveal but I was relieved that at least I’d cottoned on to the fact that I’d somehow missed the media boat and had to do some catch up.
No one got back to me.
I guessed everyone was super busy and I was being my normal over-needy self. But as an occupational pest I decided to try and find someone to help me still.
I found someone by sneaking into a VIP room, but the someone looked between the age of 16 and 22, and though trying to be helpful, she also didn’t quite seem to get what I was asking.
No one got back after this attempt, so I finally gave in.
I sat down for the main show in the second row and waited for the big announcement so I could race to get the story out before deadline, which was fast closing in at 3pm.
As things turned out, it wasn’t quite Steve Jobs January 2007 (the iPhone), although Dr Kelly did mimic the “oops I forgot to tell you one thing …” act, but it was an interesting story.
Heidi, until then a software company, was making the seemingly high-risk decision to move into hardware in the form of a purpose-built branded clinician lapel that could standalone record and transcribe consults without power or internet and integrate with Heidi’s AI software.
The story wasn’t so much the device – like Heidi’s software it isn’t complicated tech – but that Heidi was making a statement that it was so committed to utility in the clinical AI consulting that they’d decided they needed to enter the hardware game to significantly improve the recording and transcribing experience.
At the end of Dr Kelly’s reveal, at about 2:15pm or so, I was still seated in the second row madly typing as the main theatre area was noisily emptying out around me. A small band of Heidi people approached me, fronted by a young woman with a stern look on her face who, without introducing herself, curtly asked what I was doing.
“Writing.” I didn’t look up.
“Writing what?” – slightly more agitated tone.
“A story.” Still not looking up.
At this point, the young woman moved to be directly in front of me and introduced herself as the global head of communications for Heidi.
“Oh, okay,” I said, and kept writing.
She then asked who gave me permission to be writing a story.
That stopped me writing. I looked up, wondering if I had heard correctly.
I then did a full sweep look around of the room, just to check my senses were intact, and turned back to the posse, wondering why, in a room of more than 600 people with smartphones, most members of LinkedIn, I would need permission to write a story.
Remember, I had also registered as media.
I drew a breadth and asked as calmly as I could: “Sorry, can you explain that again.”
She did, this time saying that she was prepared to work with me and give me the exclusive story, with some exclusive interviews with Heidi management, if I was prepared to work with her.
Now I was confused.
I did another, this time exaggerated and deliberately theatrical 360 sweep of the room, hoping to make what I thought was the obvious point.
By now, not 10 minutes since Dr Kelly had finished the show, I had already seen three LinkedIn posts on the announcement, one from a competitor.
I tried to be polite.
“I’m really sorry but I’m on deadline, I need to keep writing if that is okay. Can we talk later?”
The global head of communications adjusted her stance and said:
“Do you really want to be on bad terms with Heidi…. because if you do you, keep writing …”
Cringing, just a little, but now with my head back down writing, I squeaked: “Okay.”
How to make a friend of and influence the global head of comms of rock star medical AI startup 101.
Notwithstanding this odd encounter, I wrote a really positive piece on Heidi’s lapel launch, deciding very quickly that the kerfuffle wasn’t at all relevant to an otherwise interesting story.
But, just to be clear, in my 40-year history as a publisher and a journalist, this was one of the more bizarre encounters with a PR and comms person I had ever had.
A few people from Heidi sent me a thanks for the piece over the next couple of days.
So why write this story now?
Because it turns out that global-comms-headgate isn’t an isolated incident.
As just one example of lots of incoming in the past month or so, just a couple of weeks ago I came across a senior engineer poached from another major Australian medical software group who left one week for a massive pay rise and asked to come back six weeks later.
Without getting into the detail, there’s enough noise from people going into and coming out of the group about disorganisation as a result of growth and speed that the seeming obsession Heidi now has with growth and speed and the possible fallout from it warrants some deeper thought.
It looks as if Heidi is performing a high-wire balancing act in its current strategy.
Which is the fine line between creative growth management and enough systemic disorganisation to mean you’ve more got a problem than a plan.
Why so desperate for growth with the concomitant barely-contained incidents of chaos that can come with that plan?
Because Heidi still isn’t anywhere near established enough to give anyone – most likely its savvy current investors – certainty that it’s going to make it.
Heidi’s current valuation of approximately $702 million is on revenues of about $70 million.
Nearly all that valuation is not based on how good their product is. It’s based on its brand, its large and rapidly expanding global user base (footprint), and its story.
If you’ve got good ideas and good developers you can make yourself a decent medical scribe today for between $50,000 and $100,000.
Heidi’s story is it started early with a “decent” scribe and marched outwards very quickly. It was literally a first mover and you could have nothing but admiration at this point of time for Dr Kelly, and for how fast it moved outwards and became a global play.
But the fact remains the product itself is pretty much a commodity not a protected piece of disruptive IT IP.
Heidi might try to spin their way out of this statement by saying, we aren’t a scribe anymore and they’d be correct.
They are racing outwards on product utility and brand to leverage their big footprint of so far only sticky-ish users, in order to make them sticky enough to have a business moving forward.
Hence, we see them developing add-on functionality as fast as they possibly can – most significantly so far, a decision support layer, which they say will compete with the current major global AI medical decision support group Open Evidence.
Add enough utility quickly enough, and you should be able to hold on to your large and still fast-growing user base.
A very sticky user base is the key. In fact, the actual goal here for Heidi and its investors is user captivity – super glue type sticky-ness. If they can get there, it’s going to be a very big deal. Something that the likes of Epic and Oracle will start losing a lot of sleep over, if they don’t just buy it at some point.
But Heidi’s strategy, despite some attempts to manage narrative and outward appearances – so far pretty bad if you think about global-comms-headgate – is still very high risk.
It has a few very sticky (pun intended) strategic problems to deal with.
The EMRs aren’t going down without a fight
EMRs all around the world, have until now been the ones who have had dibs on user captivity.
Epic and Oracle became so notorious for “information blocking” that the US government had to introduce broad, sweeping national legislation to force them and every other major US medical platform vendor to create API and FHIR interfaces to their secure cages of data, so it could flow a bit more freely for that country’s wickedly dysfunctional system.
But even so, these medical data platform monoliths still have ways and means of keeping out disruptive AI newcomers.
First, they have the money and the smarts to build their own AI for their own platforms which will be significantly superior in functionality to anything a Heidi might be able to offer a hospital or clinician who is entangled with Epic or Oracle.
They can do this because they still have vastly more money than Heidi to build such products and vastly more intel to build them to a more functional level than Heidi can, especially in terms of engaging with their own platforms.
They can also continue to do what they did in the past and make it as hard as possible for an outside piece of software, especially one so potentially competitive and disruptive as Heidi, to integrate seamlessly with their platform.
Epic and Oracle can still stuff things up here. They’re expensive and can be slow. Heidi is counting on it.
A problem for Heidi outside these global enterprise platforms is local “bespoke” patient management platform providers who are strongly incumbent.
Related
Take as a very good example Best Practice in Australia.
Though just in Australia and part in New Zealand, Best Practice’s hold on this market for GPs is a big problem for a group like Heidi.
Best Practice has about 80% share of the patient management platform market for GPs in Australia. For now, if you want in on that market, you mostly have to go through Best Practice.
Great, Heidi has an integration to Best Practice via Halo Connect. But it’s a light integration, with very limited patient data passing between Heidi’s AI functionality and the longitudinal patient record held by Best Practice.
Without the context of the full longitudinal record, Heidi is episodic for any GP using it. The GP has to do work in their head to understand their patient’s context while doing the Heidi consult, and whatever actions the GP takes afterward is a pain to get back into the data base of record, Best Practice.
In Australia, ironically unlike the US, the government so far is not mandating minimum data-sharing standards for medical software platforms. This means Best Practice can keep Heidi at arm’s length as long as it wants.
It’s in Best Practice’s interests to do that because if Heidi could interrogate the full patient record, and other important administrative data, like payments, it would be dangerous – which means amazing for the GP and disintermediating for Best Practice. With all that additional contextual data Heidi would be able to do amazing new things around a consult, some of which are the core value currently offered by Best Practice.
If Best Practice let Heidi do this, then it would be reduced to a data base function as opposed to its currently strategically strong position of an essential workflow tool.
So what does Best Practice do?
It buys a strategic stake in another medical-focused AI group, Lyrebird, and with enough of a stake to ensure they can control their own disruption journey, grants Lyrebird far more deep and seamless access to the crown data jewels deep in the system.
Although Heidi has brand, reputation and lots of evolving functionality, for now at least, it is effectively locked out of the most important software platform gatekeeper to GPs in the country.
If the Lyrebird strategy succeeds, Best Practice will be one product with one straightthrough and seamless workflow and one screen. Heidi would be out of the GP market effectively in Australia over time.
The other dynamic in play here is that while Heidi currently has an evidence and decision support layer which is making it more attractive in some ways to compete for GPs in Australia, MedLuma, developed by the CPD provider Medcast, is an Australianised version of Heidi’s evidence product which is now also integrated into Best Practice and probably will end up as the evidence layer for both Lyrebird and Best Practice.
As MedLuma has been built around Australian guidelines and content, it’s arguably a safer and more relevant decision support layer to attach to AI in Australia. That goes to trust and trust is gold in this market.
The Best Practice paradigm on a global basis
Now let’s take what could happen in Australia if Best Practice plays its cards right and look at some global markets where Heidi is operating.
Can the same thing happen to Heidi in the US and the UK that is potentially happening to it in Australia? The simple answer is yes, it can.
Heidi can’t keep trading on its “we’re a cool AI add-on to your core patient data software” persona much longer. It’s doubtful that it even can now.
Every major legacy EMR platform provider for every type of clinician has the same problem with Heidi that Best Practice has in Australia.
We know Epic and Oracle have a strategy to protect their core data platform products on AI, so we can safely assume equivalents of Best Practice in the US and the UK, like Emis and e-Clinical Works, will be developing strategy in the same way.
This might mean Heidi has to change tack quite drastically soon.
Either do stakeholder deals with appropriate EMR incumbents, like Best Practice did with Heidi, and give up equity, buy some in each big country (they will be very expensive) or bite a massive bullet and start marching rapidly backwards into the EMR market and try to take on some of these deep data incumbents with more agile AI driven PMS.
All of these strategies are fraught with high risk and cost.
A mile wide or a mile deep?
Until now Heidi has practiced a “mile wide strategy”. It’s chosen to go very wide very quickly and gotten itself an impressive global footprint.
But because medical AI technology isn’t that hard to replicate, Heidi faces another potential type of competitor: the one that decides to build themselves similar AI functionality but go deep very fast.
If we again look at Australia, that competitor for Heidi would be a rapidly evolving GP AI product called MBSPro.
MBSPro isn’t generally on many people’s radar but it should be.
It’s a young AI product (only just turned one) built by a passionate working GP and practice owner and a radiologist, with a single obsession: make workflow for Australian GPs much easier.
Without going into detail here (we are going to profile this group in the coming weeks) what you should know about this product is that it has AI working on a lot more day-to-day Australian GP data – things that either Heidi or Lyrebird do – as an example optimisation of payments and MBS coding.
The question for Heidi and Lyrebird with a product like this is, if this group can make a profit going a mile deep on Australian GPs, then how will they compete because both products so far are designing largely for all clinicians with all the functionality issues that would create if you are competing with a GP only vendor.
Heidi of course is many miles wide and compared to MBSPro and its Australian GP focus, not very deep yet.
The first thing anyone thinks about MBSPro when they come across it is, surely such a small group can’t make it in the world of big money and tech.
From what we understand, MBSPro has never taken an investment outside of its founders, is generating significant revenues already (well north of $1 million), is growing and has a user base of around 2000 GPs already.
Think about what you might pick if you’re an Australian GP today: Lyrebird, Heidi or MBSPro?
Each has its advantages. It’s not an entirely easy decision.
Lyrebird’s biggest advantage is the opportunity for deep integration with Australia’s biggest GP EMR. That provides context and the opportunity for far safer and more seamless AI functionality integration. In the end it provides more certainty around trust.
Heidi has a good product, with an evidence layer already, it puts on great parties, its CEO is an evolving medical AI guru, now complete with his own US-based medical AI podcast – Dr Tom Meets. It is trying to build a brand of success.
But it’s miles wide as far as GPs are concerned in Australia. It doesn’t have deep integration with Best Practice for that all important historical longitudinal data, and it doesn’t have the range of very GP specific data management solutions that MBSPro has and is constantly working on – at speed.
The last thing about a mile wide worth mentioning? It’s so logistically difficult to organise that you make mistakes. Global-comms-headgate was in relative terms not a huge mistake. But it is diagnostic of the issues that Heidi faces in trying to get so big so fast. You wouldn’t want similar mistakes to wash out in its engineering division.
Can Heidi outrun the storm coming?
Heidi’s chaos is perhaps understandable now.
There’s a storm coming from any which way you look at this group from a strategic point of view.
It has to grow and impress investors, because if it’s going to disrupt its way into the incumbent medical data platform markets across the world, it’s going to need a lot more money and fast.
It doesn’t have a product that any of the smart and cash-rich incumbent EMR vendors can’t reproduce themselves, or buy or partner with.
AI as applied to patient data through a clinician patient management EMR or dashboard is not a strategic moat of any sort.
This means Heidi will need to buy, deal or bash its way into the core patient data EMR market with its current brand proposition: agility, speed of functionality development, and its existing brand connection and footprint.
When I heard that Heidi CEO Dr Kelly was getting his own US medical AI podcast, featuring an expensive producer from Harvard, I initially wondered what is he thinking … doesn’t he need to spend his time on product?
The answer is probably no he doesn’t. He needs to spend his time on investors and vision.
He may have lost Australia already, unless of course his investors decide to buy that 51% of Best Practice that’s apparently on the market.
But Heidi’s mile wide strategy makes that a bit of an expensive decision.
It’s going to be an interesting few years to see how this all pans out.
