The Sifted Summit 2023 brought together some of the brightest minds in the startup and tech world to discuss the latest trends, challenges, and opportunities. Here are some of my takeaways from this year’s summit:
1. Bridging the Gap Between AI Ideas and Reality
AI – and especially Generative AI – was one of the recurring themes of the summit. On the one hand, there were mature AI companies (usually speakers, or service providers supporting start-ups through AI-based offerings) who talked little of their AI, and far more about what value they could bring. For example, I met one CEO who would sell you a subscription service to produce reports, forecasts, or sales strategies based on similar companies but using your own data. The idea being that you could replace some of your C-suite executive functions. On the other hand, there were a lot of start-up CEOs (almost everyone I met was a CEO) who didn’t know where to begin: they needed consultancy to help them begin their AI adoption. And no-one knew much about prompt engineering
2. Generative AI: Transforming Industries
I met some interesting start-ups that were using AI to change their industries. For example, wearable tech wired into diagnosis and prediction models of disease; creating more realistic rendered images for architects’ drawings; or simply replacing much of the document generation currently done by lawyers and other professionals. What was interesting here was the feedback loop: how do you make your models better and prevent drift away from what was correct, aesthetically pleasing, or legal, depending on your use case. I heard about using different AI technologies to check each other, even different versions of the same technology, with the more stable technology asking consistent questions and assessing responses, with the newer technology providing those responses.
3. Forming Better Questions with AI
Generating quality responses from proprietary data through generative AI models was another hot topic. It’s not just about finding answers; it’s about asking better questions. The goal is to provide natural language feedback to users, making data inquiry more accessible and intuitive. For example, your users might interact with ChatGPT, which then reforms their question, passes it via some intelligent routing to your own data and documents, e.g. where should I look for HR processes, or sales data, or corporate CO2 emissions data, and then collating the presenting those responses through ChatGPT again. Elasticsearch seemed a useful choice for the back-end data as it could be pre-indexed according to your own organisational knowledge.
4. The Startup ecosystem
There’s a whole ecosystem of supporting organisations that help start-ups in the early stages, whether that’s for accounting, recruitment, contracts, sales, and part-time or “fractional” C-Suite. We happen to use on of those organisation (Deel) for contractors where we don’t have a legal entity. It’s now easier to create a start-up and buy in these services until you can afford to run them in-house.
5. VC Perspectives
There was a general view that this is a challenging time for growth-stage start-ups, and that this is will last for at least another 6 months. Founders have had to shift from capital-fueled growth to a more product-led approach. This has meant reductions in staff (40% was suggested by a few), with some founders even returning to the drawing board to rethink their product/market fit. VCs are also now expecting metrics focussed around unit costs and profitability, rather than growth projections.
6. Start-up Advice
There were some useful insights for founders on how to manage companies once the initial idea has been tried out. The messages I took from this included:
- Tell your employees what’s going on and why. Story-telling is often a key part of this, and those stories need to be demonstrably true
- Employees are much more interested now in work with a purpose: it’s not simply a transaction of wages for time/effort,
- Ask yourself the “why” questions first: why does your company exist?; why is your product any better than anyone else’s?; why should customers buy it?
- If you make claims about your company, can you back that up with your processes and polices? For example, “we’re a learning organisation”. Great! Show me the policies that guarantee a budget for employee development; the process for selecting who goes on which training course; how that learning is fed back to other employees; evidence of “lessons learned” exercises after each organisational failure or success.
7. Data origins and applicability
We’re probably all aware that data biases exist: car crash test dummies (introduced in the 50s) were based on the 50th centile male. Not so great if you’re making claims about how safe your cars are for female drivers. Many start-ups are keen to layer their own value on top of publicly available data and perhaps aren’t scrutinising too carefully how applicable those data are to the start-up’s intended customers. We heard from one founder who had identified this and then sought more representative image libraries upon train their AI models. In her case, the bias was obvious: it was showing a high proportion of muscular white women, where her target customer base was diverse. For other start-ups who were not using image data, these biases are hidden and may not even be available: do the data relate to men?; to people from the south of England?; to the wealthy or educated? These founders weren’t necessarily trying to conceal this, but needed to ask themselves the right questions.
The Sifted Summit 2023 was a treasure trove of insights, sparking discussions on the challenges and opportunities in the startup world. As the tech landscape continues to evolve, and evolve fast, these founders will need to navigate these themes and take advantage of the new technologies to reach their markets