I worked for 3 years on a startup product worth $500 million only to watch it die a slow and painful death

Startups have occupied a good part of my life. In 2017 I decided to bet everything into starting something I believed had potential. Elisebot is that startup, and this is its story.

It’s a story filled with high highs and low lows. It’s a story of struggle and passion, despair, and helplessness. It’s the story of a great idea, a good product, a competent and diverse team that defied the typical founder profile. It’s a story of a company that almost made it so many times, only to fall flat on its face. Now that it has tumbled its final fall, this story shall be shared with you, dear reader, and the rest of the world.

My startup journey

Since 2008 I have been actively involved with the Lean Startup movement. I read every Eric Ries blog post, I’ve listened to Steve Blank talk a million times. I have read almost every book written on the topic multiple times. I interacted with fellow advocates on forums and mailing lists.

I have tested out at least 5 different product concepts, advised and/or worked at 3 venture-backed startups, 2 of which had exits and I haven’t seen a single share of. I participated in multiple startup events, hackathons, and startup weekends.

After working at Kidux, a child internet usage monitoring and parental control application in 2009, I moved with my wife and our son aged 5 from Brazil to Belgium to join Belgian startup Trendminer as software engineer #1 and potentially become a partner.

I worked with them during 2010–2012, planning and building their first product from the ground up, developing, hiring, and managing a software engineering team of 6. Together with the current CEO, Bert Baeck, who was then Head of Sales, and our development team, we delivered the first successful implementation project to one of the top 20 companies in Belgium in the process industry.

I never got the shares I was promised, and after being micromanaged out of my job by the ex-CEO Noël Jans, I left the company and started looking; they went on to build a second product and got acquired by Software AG last week.

After TrendMiner, I helped friends with their own startup ideas and tried a few of my own: StartItUp, which was basically Co-founder Lab before that existed, Convify — a conference social app to help you explore “the hallway track”, YourStage — a platform to book musicians for your event or venue, among others. None of them got too far, I always struggled with customer validation, being stuck in Belgium and trying to do B2C things for mostly the US market. You can’t underestimate how hard it is for a foreigner to start a business in any European country — people like doing business with people like them.

I also joined YC alumni Teleborder in 2013 for 3 months, worked on a startup idea with some college grads in the US, inthis.co, and helped define the initial architecture and implemented the storefront side of payments processor Sign2Pay in Belgium. None of these worked out — people seemed to be making the same mistakes I’d seen happen before at TrendMiner or that are well documented in posts and books in the Lean Startup community.

They all hired mindlessly, adding in anyone who was willing to work for them for a low salary, rather than waiting until (or even past) the moment the current team was overloaded with work to hire. The marketplaces were always focusing on the provider side rather than the customer and ended up failing to get traction with customers. They were all building WAY too much product before there were customers, and before they had Problem / Solution fit, let alone Product / Market fit.

I figured if they won’t listen to my advice, I’d rather find someone who did.

I went to Startup Weekend Brussels in 2011, worked with some great people on a food delivery app way before Deliveroo or Uber EATS were popular, and kept on working on that for a bit after the weekend with one of my teammates.

Later we applied to YC with AirJudge, a cloud-based workflow engine that was programmer-friendly and easy to implement. That didn’t go through, we never got even our rejection e-mail. My partner went on to join a dashboard startup in London, and then became CTO of a fintech startup in the Barcelona scene. I kept on looking for the ideal co-founder.

By the end of 2016, I met a great guy in Brussels who wanted to create an app for guiding people through scenic routes during holiday road trips. We had dinners and lunches, worked on strategy together, settled on making a bike routing app first to try out the collaboration.

I wasn't able to hold up my end of the bargain very well. He managed to do a lot of customer development, and I didn’t get to do as much software development as I thought I could manage due to a job change. We went our separate ways.

I then started working on a chatbot, and the rest, as they say, is history.

The beginning of Elisebot

Fascinated by the possibilities of applying this technology to the enterprise space — I’ve been working on enterprise software for the better part of my life in some capacity — I proceeded to collaborate with a consulting client in creating a Slack chatbot that offers up to the minute information about orders in the Odoo ERP (previously OpenERP).

After a trial with wit.ai, I realized their product wasn’t mature yet to implement anything — results were random, unstable and the whole thing mostly just threw errors with every call. After switching to using api.ai (now Dialogflow by Google), I got the proof of concept working nicely.

Later, when shopping this idea around to friends and clients in my network, one thing became as clear as water: CRM adoption was a BIG problem, one that natural language interfaces are VERY well suited to solve. I have built 4 custom CRMs from scratch and worked on implementations of other CRM like Salesforce and Microsoft Dynamics CRM over the years.

Salespeople are always on the road, usually driving from one client to the next, and don’t have time to update their CRM, effectively driving their managers crazy. Managers are seriously burdened by having to call their people every single day to remind them to fill in meeting notes and set up follow up actions in the tool.

With that in mind, and problem/solution fit guaranteed by doing customer development in my own network, I recruited the best UX designer I know — my wife Carol— and we got started on building the prototype of what would become Elisebot: a voice-driven AI agent that you talk to update the CRM from the car. The prototype could record meeting notes, schedule follow-up actions, and manage the stage of opportunities in the pipeline.

Prototype and the first accelerator

Recording a team video was an exhausting effort for both of us. People running accelerator programs assume everyone is comfortable in front of a camera and was born taking selfies — not quite so! Me and Carol are camera shy and try to keep our digital footprint to a minimum. It took us 2 whole days of work to land a take we actually felt was at least OK.

After a great first interview with their team, and having them try a demo of our prototype, we had a second one scheduled with Gizmodo’s one and only Peter Rojas , who is also a partner at Betaworks, and Betaworks’ CEO John Borthwick.

I don’t know if it was the blizzard that turned their lunch break into a harsh walk (we were up right after lunch), or if John B. didn’t believe at all in the idea but was talked into interviewing us anyway, but we were met with a very skeptical CEO, asking us “how would we get every salesperson in the world to use our product”, and is very dismissive of anything we said. Peter, on the other hand, was the most friendly person in the world. We are thankful anyway for the chance of pitching to them and have carried on working on our product. In the end, it was probably for the best, Betaworks’ interest lies in reusable viral things like Giffy and consumer internet things like Digg, not a great match with an enterprise software startup.

At this point, we had almost eaten up our life savings due to me not working on my freelance contract to work on Elisebot in order to finish the prototype before the application deadline for voicecamp. On the other hand, getting this far with just a prototype totally lit us up, and I spent the next month trying to get through the door of every potential client I could get a grip on. I e-mailed and I called and I reached out to my network, only to be met with a very loud silence. We kept on working on turning the prototype into a product.

Early hustle

I managed to get one of the clients I e-mailed to respond, a Point of Sale (aka cash register) equipment brand from the UK. Their COO was really nice to me, keen to help a startup out, and put me in touch with their CRM manager. After weeks of rescheduling, I managed to talk to him. He wanted to show me how we could best use Salesforce at Elisebot (maybe the COO got the wrong message?), and as far as our product went, he thought maybe their post-sales team could use it, but the timing was off, so we postponed to April. Further contact never even got replied to, until some point in 2018 when I asked harder and they said they were no longer interested.

I reached out to the CEO of TrendMiner who was excited when he heard about my idea for Elisebot and set the wheels in motion inside the company to have their sales team try out our prototype. Their VP of Sales was really enthusiastic about having their Account Executives try out our solution, and we set up a pilot project with their salesperson in the Netherlands.

I skyped with their rep once, showed him the ropes of Elisebot, and asked him to try it out. I followed up on him every week for a month, and he still hadn’t tried it out a single time. You see, the problem is Elisebot was built based on the assumption people keep their meetings in their calendars, and that their calendar is synced up with the CRM, but we quickly learned it isn’t quite so. My Dutch friend found a chore to have to sync up his calendar manually.

After trying to push the Dutch rep to use the solution to no avail, I decided to try out their other sales guy in Germany. He was already using a better interface to their Salesforce CRM, ZeroKeyboard. He wasn’t really keen on trying Elisebot, and what he really wanted was an iPad Pro, which it seems his manager thought was not necessary. Oh well. In retrospect maybe I should have bought him one in exchange for him at least trying our product out.

So I waited for their new rep in Houston, TX to come on board 2 months later, as their manager said she would be best suited to try it out due to being a native English speaker. Mind you, at this stage, there wasn’t a single interaction with the product, nor in Native English, Dutch accent English, or German accent English.

I tried contacting the rep and she never replied. I contacted the VP Sales about it, but he was always too busy to talk. The CEO was busy trying to negotiate a capital rase or exit. Seems like we didn’t have a pilot project to begin with.

Hustling is hard

I managed to give the EdTech company a demo, we scheduled a new meeting within 3 months as the CRM manager was still new to the company and was focusing on other projects, and down the line, in the day before our meeting, he canceled citing other priorities.

The university in Australia just wanted to check the product out, but they didn’t even have a CRM yet and had been trying to decide on which one to implement. Consultative selling to help them in that choice brought us absolutely nothing. Past the initial phone call with the CEO, he never even replied to my e-mails, even though I had spent a good 12 hours helping him choose a CRM.

At this point, we are sleep deprived, our house is a mess, our kid is not doing great at school, and our savings ran out. It was time to face the harsh reality of having to park Elisebot for a while — at a huge opportunity cost. We couldn’t know how big it would be, but there was definitely one and we could sense it.

The second accelerator

Following a nice evening where they explained the program to some 20 people who showed up to the event, we headed for cocktails, and after talking a bit to Rodrigo, I left with the impression they were convinced we had a good product/idea in our hands and were capable of executing on it. A month down the line, all I get is the standard TechStars rejection e-mail. Go figure. A trip to London and another €500 down the drain if it wasn’t for the fantastic late-night barbecue and chat with a friend, who was really patient waiting for me to be done with the cocktail, and the best company I could have asked for.

The 8-month long first date

I meet with him, and we hit it off. I explained to him the struggles we went through, and how we kind of saw this evolving into more of a consulting + product model, which matches the consulting focus of this conglomerate of over 400 companies.

He invites me to pitch to a larger audience comprised of many of the companies inside that group. They would evaluate the product, idea, and decide if they would be interested in taking any next steps.

I was nervous as all hell, pitching from a stage to a very skeptical and unfriendly tech-savvy audience who all though they could build the same thing in a week (turns out they can’t, but hey…)

Anyway, the demo wins them over enough to get me invited again, this time to discuss the terms of a “trial collaboration”. One of the companies in the group, a Salesforce integrator, offers a partnership to co-market our product to their clients.

Since that first pitch, I start drumming up the sales drum again. Trying to find local clients, trying to reach out internationally, anything to show some traction to get this deal going. Nothing.

I buckle down and make the prototype airtight, fix every bug. Carol tests it to exhaustion, we were really keen on having it be the best it could be in order to raise capital and get clients. That conglomerate has a sales team selling everything in their portfolio to the biggest companies in the Benelux.

Next meeting, I and the conglomerate's CEO sit down for not even 10 minutes. He decides they will lend me 2 people for 15 days to work on the product, in order to figure out whether we are a good fit for working together. We would get the product ready for the Dutch-speaking Belgian market while leaving things open to working in French as well.

The co-marketing partnership gets a kick-off too, and we promise to put a video and new website together. That would go on to cost us a month of work and about €1k between stock footage, voice-overs, and tools we bought to edit the video, not to mention the time and money lost from consulting work.

After an overwhelming total of 4(!!!) kick-off meetings in the span of 2 months, we get started working on the product with 3 people to get it translated and internationalized. On the very day of the last kick-off meeting I’m catching a flight to Brazil, my stepfather had been brutally murdered while trying to stop car theft.

I’m back after a week, and 3 weeks of hard work later, the project is a success. We finish it off with a demo to the CEO. The CEO is convinced of our strengths as a team in Elisebot, and of our product (or so I thought?). We discuss the terms of their investment and decide on the next steps. I would speak to their Legal department to get a company incorporated, transfer the Elisebot IP to the new entity and we would be in business come February with enough money in the bank to last us a year developing the product further.

First commercial breakthrough: TC Disrupt

We walk to the venue, set up our stand. I speak to investors — lots of attention, but all of them want traction before putting down their money. Our stand is visited by some Fortune 500 companies, and we walk away with at least 10 good leads. George did the rounds on the other stands scouting for partnerships.

After the show (we didn’t compete in Startup Battlefield in case you’re wondering), I call those leads, manage to schedule demos. Out of the 10 leads, only 2 of them are interested enough and have the right timing to begin working with us. Contracts go out in February — no contracts signed yet as of July 2018.

The Conglomerate Reloaded

I start exploring other funding options: subsidies, angel investors. After meeting with the Flemish Agency for Entrepreneurship, we come to the conclusion the only viable path for us to raise subsidies would be through an innovation project. Peter Rutte, one of the coolest guys I’ve ever met in the Belgian startup scene, helps me write a project proposal.

Peter also lands me connections within the University of Ghent, and after meeting with professor Kris de Muynck, it becomes clear that the challenges we were currently facing with the product were non-trivial, and they would definitely be open to collaborating with us to create patented NLP and AI technology to improve our product.

There is a small gotcha with the innovation project subsidy though: they only supplement your own capital. If you put in €100k, they put in €95k in subsidies. The problem is all of our savings had already gone to pay the bills while building the product, so we had zero euro. Everything hinged on whether the conglomerate would go through with the investment or not. I asked the investment arm of my bank, ING, but they said the usual “you’re too early for us, but let’s stay in touch”.

I sit with the conglomerate's legal team and my mentor inside the company, and we come up with a new investment proposal: cutting off the consulting arm and focusing on the product. We also cut the investment amount in half for the same amount of shares. The legal team tells me the CEO will get back to me, and after following up with them and the CEO for 2 months and trying to get another meeting scheduled with the CEO in vain, I finally realize he’s no longer interested in investing and decide it’s time to move on.

At this point, I’m frustrated with the startup ecosystem, and after watching Michael Seibel, Y Combinator’s new CEO after Sam Altman stepped down, give talks and say they are focusing on diversity, I e-mail him, asking whether they would be open to considering an application from a 30-something husband/wife founding team. He replies:

we have no problem with married founders of any age and have funded a number of couples in the past (that is a strange thing to be against)

you should apply!

And so we do apply. We got the rejection e-mail late at the final decision date. What a surprise.

Contracts are not sales

Turns out they have a lot of industry-specific vocabulary that our speech recognition engine can’t handle very well, definitely not when those terms — in English — are within a conversation in German. We promise them that we will work on a custom speech recognition model, but ask them to commit to buying 3 users for a year. They back off and ask us to let them know when speech recognition is improved.

The other lead — an EU organization —looks promising. After 2 months of conf calls with their team, they’re supposedly ready to sign — until they discover the CEO won’t give them autonomy to decide to use our tool, and they have to run through their European Union bureaucratic purchasing process. Our internal sponsor loses interest in the project, and the contract is still lying somewhere in the CEO’s desk to be signed. Hopefully I’ll hear more by the end of the summer. All I got was a GDPR compliance grill from their IT security person.

The timing finally matches up for a big courier service from Germany to start evaluating our product. I had a meeting scheduled with them at the end of May. The day before, they ask to reschedule to 1 day later — sounds promising, at least they didn't cancel! During the call with their business developer, I get email introductions to VPs inside 2 internal divisions. After trying to reach out to them and an initial good tempo, the dark vision of the summer holidays looms over and everything starts getting delayed. Not even a single phone call with the VPs or their teams goes through.

The last accelerator

After going through the heartbreak of applying to Betaworks’ voicecamp and Y Combinator S18, I didn’t feel like going through it again. Lo and behold, it seems someone out there was interested after all: we got an email from a pipeline scout for the SAP.io program saying they asked the program manager for a deadline extension and inviting us to finish our application. So I did, re-used the same team video from 2017 — didn’t want to go through all that pain and shame again of being in front of a camera trying to explain why we’re so awesome in less than a minute; sometimes these accelerator programs feel like a circus with these arbitrary video duration and paragraph length rules. I sent in the application and tried to forget about it.

We got invited to interview for TechStars Paris instead of SAP.io (left me scratching my head after we’d been scouted by the latter), and so off we went through the emotional roller coaster of accelerator interviews. This time around, the interview went really smoothly. We spoke to Julien, the program manager for Paris, and his assistant Queta Gonzales. The questions he asked us were spot on, and he wasn’t at all skeptical. We got cautiously excited.

A few days later, they let us know we’d be going through a second interview with the managing director of TechStars Montreal, Sunil Sharma. At the scheduled date and time, there we were in our conf call. Sunil was a little bit in distress and on the go — he was driving, which made perfect conversation out of Elisebot’s intended audience — but thankfully he was a gentleman: he switched off the camera, and we had a really cozy, nice chat about our background, the product, potential uses, and he even suggested one of his alumni companies for us to try and partner with.

Cue a week of emotional roller coaster/e-mail refresh madness, and by the very end of the Friday we were supposed to get a reply, they apologized for delaying it until the next Monday morning.

At precisely 12:20 PM on Monday I was writing to ask whether I was supposed to remain available, and right after I hit the send button, an e-mail came in saying we weren’t selected for the third and final interview. Guess there was no last stint of hope after all.

Time has run out

It is extremely disappointing to think that after searching on a weekly basis for competitors for 1.5 years, I failed to find Tact. They were further along, their product had a more mature feel to it (even if our product video looks and sounds better), they had ties with Salesforce, and they were US-based. Boom. Cue the world crashing around me in slow motion.

As a final blow, their slides boasting about current clients showed Medtronic’s logo loud and clear. I guess I know now why their IT Director didn’t reply to any of my e-mails.

With a capitalization of over $50mm, it is mighty hard for us to stay competitive with them. We didn’t manage to raise any money. They raised $50mm. That puts their valuation in the $500mm range.


Surely enough, Tact's CEO has a story of working for Siebel and Salesforce, making him a more suitable candidate than me for leading this effort, but still. Why would no one even let us in? Because VCs invest in teams.

The truth about “we invest in teams”

VCs like to invest in people they think will be successful. They tout “we don’t invest in ideas, we invest in teams” — which makes one think everyone has a chance if they’re smart — but in reality, it means something completely different.

Here’s what it means:

  • If you never had a successful startup exit;
  • If your company doesn’t have more revenue than you need for funding;
  • If you have not been a high ranking executive in a Fortune 500 company;
  • If you haven’t worked for Google, Facebook, Apple, or Amazon;
  • If you don’t have an MBA from Harvard;
  • If you didn’t graduate from a CS class at Stanford;
  • Even if you graduated from MIT, but didn’t make any good connections
  • And if it’s been more than 2 years since you left college and you haven’t done something amazing yet

You are shit out of luck. You’re not the right team. You don’t match the pattern VCs expect from a “smart young person” that can take a startup from 0 to unicorn billion.

Don’t trust me on this? Read away:

We had the right idea. We had built the product, and it worked. We had contracts sent out and a decent indication of traction. But none of that mattered.

In reality, what all those accelerator programs and investors saw were a 30-something Latino (arguably Brazil is anything but Latino culture, but whatever) couple, living in Europe, with a kid, who didn’t at all match the pattern.

The sad truth is Venture Capital, and startup culture, in general, worships the “smart young kid”. The Stanford grad dude with an AI masters teamed up with the cutthroat Harvard MBA guy. And since VCs invest “in the team, not in the idea or the product”, this won’t ever change.

Next steps

I’ll be serving as a sounding board for the Nimona project, a peer-to-peer network built with privacy and data ownership as a center piece from the ground up. Yes, like Richard’s new internet from Silicon Valley. It will give people back the control over their personal data, enable multiple applications to benefit from a single version of your contact network and profile, and allow people to keep their data in the cloud without handing anyone control over it. It’s awesome! I might convince them to apply for YC as a non-profit.

Other than that, I and Carol want to get even more experience in launching products. Rather than going for big ideas, we’re gonna try several small things, try to optimize for learning, and explore spaces we find interesting: plus-size lifestyle, the management of handmade businesses, and more.

Whatever it is we are doing, we want to bootstrap it all the way into profitability. The VC ecosystem chews and spits out people like us, stealing ideas, pretending to care, and in the end, never taking real risks.


  • Carol Santello, my wife, for accepting the challenge of building this with me and going together through the emotional roller coaster that is starting a startup;
  • George Antoniadis for doing way more than I could ever have asked of him. You coming to Berlin with me was surreal man, coding together was a blast too. I have no words for how thankful I am for all your help and encouragement with this;
  • Bert Baeck / Bert Baeck for believing in me at Trendminer and bringing me to Belgium; for helping me in getting started within the Belgian startup ecosystem; for agreeing to be an advisor on Elisebot once we got funding; for helping me think the business through, and helping me every time I got stuck;
  • Aziz Ajaji, Alastair Woolcock, and Anthony Meulen at Gartner for the mentorship through trying to raise capital during the pandemic and lifting this ship off the ground;
  • Jeff Haynie for willing to mentor me through the difficulties of balancing hustling and hacking as a first time CEO, and for taking the time to hear me out even if we’ve never even met in person. Jeff sold Appcelerator not too long ago, and Vocalocity before that, and next to being an inspiration to me as an entrepreneur, is also an awesome human being;
  • Peter Rutte for believing in Elisebot, helping me write my VLAIO project, and introducing me to people who would’ve been crucial for the project’s success at the University of Ghent;
  • Frank Poncelet for helping me navigate the Cronos labyrinth, following up on the project when we joined forces and trying to help me get the investment deal there through;
  • Ine Meeusen for introducing me to Jef, and for starting what could have been the springboard Elisebot needed to reach success. You did everything you could and I truly appreciate it;
  • Aagje Reynders for her willingness to understand the maze that was the Elisebot prototype code, and making our joint project a success;
  • Ruta Stasiunaite and the Salesforce team at Quandoo for putting in the effort to gather the requirements we needed to make a convincing demo to their sales team, and collaborating with us in the best way;
  • François Xavier Fanard for believing in Elisebot even in its embryonal stage, and pushing through barriers inside his organization to enable them to help Elisebot, even if we don’t fit the profile of companies they work with;
  • alexandre mobrige for believing in Elisebot and willing to have OpenText use it internally, even if it never came to fruition;
  • Last but not least, Jim Myhrberg and the romdo collective for the support, the listening and the encouragement. For celebrating the high highs with me, and pushing me through the low lows, and reviewing this ridiculously long post.

And I would like to thank you, dear reader, for reading this story to the end. I hope this has proven useful to you in some way.

I would love to hear your comments. If you have been through a similar experience, I encourage you to share. Stories of startup madness and failure don’t get shared enough. Stories of how founders are shunned by VCs and accelerators, and how it’s not at all about building a great company.

You can follow me on twitter at http://twitter.com/alexmreis , and through Medium.

Co-Founder & CEO of elisebot.ai