Changing Engines Mid-Flight: Product Evolutions

When I began my career, the playbook for enterprise software was as follows: 1. Build a bespoke solution for a customer willing to take the risk (out of burning need); 2. Rinse and repeat a few times; 3. Generate re-usable IP that is accompanied by a healthy dose of custom professional services; 4. One day you call it a software company. You sold a perpetual license to the product, and charged an annual maintenance fee. New versions of the product got released every two years.


In the Software-as-a-Service (SaaS) era, here’s how software gets built: 1. A founder(s) builds an MVP and finds customers willing to pay for it; 2. Customizability gets built into the product, as part of a toolset or through open APIs and SDKs; 3. The product is hosted in the cloud; 4. Rinse and repeat. The product gets bought for a monthly or annual fee, creating “recurring revenue”. Companies improve their product through iterative development, often on a near-daily basis.

The SaaS approach has won the battle for reasons that are well outlined here. Transitioning between these two models while maintaining growth and capital efficiency is one of the most difficult things to do in software.


Inbenta, an Amasia portfolio company led by Jordi Torras, has done just that: redone its business model and the product while still growing at a healthy clip and without hemorrhaging cash. 

Inbenta provides sophisticated natural language processing (NLP) software for online customer support and commerce. It is a market leader — the company takes on Microsoft LUIS, IBM Watson, and Google DialogFlow, among others, and wins (often).

The Inbenta story is more nuanced than a “license software to SaaS” transition. The product was never sold as a perpetual license, and it was already cloud-based. The shared application was the NLP engine (and associated services and dashboards), and every customer had a tailored user instance on AWS. This provided flexibility and functionality, but the costs in scalability and maintenance were significant.

This state of affairs reflected the history of the company: a highly customer-centric approach that began in Barcelona, with no venture capital, applying meaning text theory (MTT) for demanding customers on a bespoke basis. Jordi and his team had no choice but to build out the product via these customer projects. 

By 2016, when we invested in the company’s Series B (and a big thank you to our friends at Level Equity who led the round and invited us in), it was clear that the product’s architecture was going to be a hindrance to the company scaling at a rate commensurate with its potential. 


Here is what the company did (this is a compressed version):

  • Precisely define its product suite (“this is what we do, and this is what we won’t do”)

  • Create APIs and an SDK that sits on top of each product

  • Get much more formal about the product management function — not just roadmaps, but a substantial re-jiggering of how product gets built

  • Document, document, document

  • Redeploy talent to triple the size of the engineering team and greatly shrink the professional services team. 

  • Have the entire team (including sales, support, customer success, and executives) provide a consistent and authentic message to all existing customers

  • Create a migration plan for legacy implementations

  • Rewrite sales material and, more importantly, retrain the sales team 

  • Train support teams to migrate legacy customers and help new customers use the new self-service features of the product

Twelve months later, 100% of new instances are based on the “new” product, and 50% of existing instances, have been migrated. We’re talking about over 250 customers, in 30 countries, on five continents.


This is really hard to do (especially while still growing nicely and not setting fire to piles of cash). This is not a high profile company with multiple boldfaced announcements in TechCrunch; this is a humble team led by a humble founder that just gets things done.

I have perfected the VC skillset of somehow claiming I know it all whilst knowing very little. The reality is my management teams teach me both simple and complex lessons every single day. I’ve learned much from the Inbenta team about making tough decisions, pursuing excellence within the constraints of real life, and simply getting on with it.