Author’s Note: This is an expanded version of the May edition editorial of our monthly Insignia Insights Newsletter.

The letters on everyone’s lips these days are “A” and “I” or “GPT” if they’re feeling fancy. While the “how to use ChatGPT” Linkedin posts (including ours) and generative AI apps and tools have exploded into the mainstream, there remains a lot of untapped potential for AI use cases, especially for enterprises.

It’s only been three years since the pandemic forced many businesses to undergo varying levels of digital transformation, and now they are reckoning with another tectonic shift: AI transformation.

Even though it’s still early, it is clear that AI adoption — and that goes beyond GPT or LLM use cases — is at a critical inflection point, changing the way companies are built and what startups optimize to find product-market fit and growth. AI tech is not new for sure, but over the past year, tools to leverage massive data sets have made new layers of use cases more accessible to businesses.

For a venture capital firm like ourselves, the question is how will this change the “rules of the game”? What will the enduring companies of the future look like?

1. Forget GMV/GTV, the real game is around data. “The companies that will do well are companies that have a system of record, that have enough data within their product set, either from a product standpoint or from a user standpoint, and utilize that data to make the user experience better,” shares veteran Japan SaaS investor and founder of AI-driven SaaS benchmarking platform projection-ai Shinji Asada on our podcast.

And we already see this approach pay off for companies like AwanTunai, as they have focused on developing their unique data sets and proprietary risk management to drive growth (as opposed to the other way around).

“A lot of the e-commerce marketplace folks have scaled up their systems to optimize for GTV capture, and sometimes there’s a bit of a perverse incentive where if there’s some fraudulent GTV, it’s still captured because it’s driving up the valuation. But for us, since we started life off as a lender and risk management is part of our DNA, we custom built the system to detect fraud,” shares CEO and co-founder Dino Setiawan on our podcast.

While some companies are not just building their own data sets but enabling other companies to unlock the value of their own fragmented data operations, like bluesheets.

“So the first piece being we provide our solution that actually does the data processing provides the automation and efficiency gains. There’s a really tangible bottom-line saving there and an improved customer experience, but following that, that organization then has a really robust data set at their disposal,” shares COO and co-founder Clare Leighton on their latest podcast with us.

Then there are also companies that are focused on validating data flows, for example, customer verification, as in the case of Verihubs. For these enabler and infrastructure applications, a key moat will be to secure certifications and regulatory recognition.

As Verihubs COO Vivi Mandasari shares on our podcast, “…we have received ISO 27001. That’s regarding the improving information security and management system, and also we have the certification for our AI. It’s called an NIST-FVRT [certification]. In participating in the NIST-FVRT certification with success in entering the category one to-one verification to many identification and presentation tech detection. With these three certifications, it proves our technology can become a reliable service in Indonesia and the world also.”

Finally, there are companies targeting the application layer of this data value chain like WIZ.AI developing solutions on top of their AI-native customer engagement technology for companies to leverage large language models and OpenAI’s GPT APIs to rapidly design and execute omnichannel customer engagement campaigns. These application layer companies are essentially building AI co-pilots for specific functions.

As WIZ.AI CEO Jennifer Zhang shares in the official press release on the launch of their first LLM for businesses, “The future of work is one where AI acts as a co-pilot, working seamlessly alongside humans. With our bespoke large language models, businesses can fully embrace AI’s potential, revolutionizing their processes and propelling industries forward.”

This maturing of the “data value chain”, especially in markets like those in Southeast Asia where data operations remain largely fragmented and disparate, will have ramifications for the kinds of companies that will dominate the market in this decade.

2. The premium on cybersecurity and data privacy as risk mitigation for growing organizations will only rise. Cybersecurity and data privacy as foundations in company building becomes more important than ever before as data becomes central to any company’s sustainability (check out our public access website security scanner tool). For generative AI tools, this has led to a race to enable or build customizable or vertical large language models (LLM) for business. WIZ.AI in particular recently launched their first LLM for business and have written extensively about generative AI, including how to mitigate data security risks.

3. More robust risk management and reporting on top of “data value chain” will raise standards. More available tools to ingest, process, and leverage data sets (be it unstructured or structured, offline or online) will make it easier for companies to do risk management and sustainability reporting, especially within the context of the ESG framework (check out our upgraded ESG assessment tool). For example, data processing and workflow automation platform bluesheets recently made ESG reporting more collaborative and accessible through their mobile app. This could potentially impact the way corporate governance is done in terms of the ability of a board and/or management to monitor key data to their organization.

bluesheets CEO Christian Schneider adds on the podcast: “…we have seen that the data we process and the data we bring to life can be used in various different ways. The one-to-many approach in terms of sharing that data, maybe with an internal legacy system works just as well as sharing it with an external database…Since we’re pulling together data from many different sources within the company, it’ll open up new use cases for the company to do ESG reporting.”

4. Companies will be expected to have data-literate workforces for more robust data operations. The shift in focus on the aspects covered above (building quality data sets, cybersecurity, data privacy, risk management and reporting) means the future of work is one where data literacy and operations will be increasingly prioritizing as a key skillset or requirement for workforces.

While it’s exciting to see how AI is changing the meta for company building, it is clear that the “devil” is in the data (and as a corollary, who has access to this data). In other words, how effective or sustainable an AI engine will be is only as good as the data it is fed and how this data is fed.

“My chief science officer once said that artificial intelligence can’t beat a trained human…Deep learning models need tens of millions of data points and users, and only a few companies have access to these data sets…​​understanding risk management is critical before building science models that can surpass human performance. And to get the right set of variables and proxies, experienced humans need to guide AI engine development,” Dino also shares on our podcast.

These are just some implications of the accelerated AI development we are seeing these days and there are more worth discussing (impact of reinforcement learning ceiling on B2C vs B2B adoption, lowering cost of building vs rising cost of selling, generative AI technology reaching the point where it will draw regulator scrutiny).