DATA DATA DATA – FEBRUARY 2024

Do you want to grow your business, innovate, drive efficiencies, &/or create new revenue streams? Of course, you do! Well data is the starting point…and this is the only data-focused roundup you’ll ever need!

In this issue, it’s all about exploring if your organisation is ready to harness the power of AI and ChatGPT-style technology. The rapid pace of technology evolution is exciting and inspiring…but you could very easily feel overwhelmed by the opportunities presented. As I shared in our December issue , there are thousands of tools available in the market, and more launching every single day, so feeling overwhelmed is entirely understandable. So, let’s see if I can make it simple for you.

Is your organisation is ready to harness the power of AI and ChatGPT-style technology?

The rise of ChatGPT has been unprecedented…

…but how you do you reap the rewards whilst minimising the risks?

We need to get our organisations ready to truly harness the benefits of this fast-evolving technology. By ready, we need to think about:

  1. Our data & AI strategy – this needs to power our business strategy and help us achieve our goals…whilst also realising that data, AI and technology can inspire our business strategy too. We need to be clear why we’re doing things, when, where and how. This doesn’t need to be some 50-page document…in fact, if we can describe it succinctly, we’re probably doing well. So, what’s your data and AI strategy?
  2. Our opportunities & use cases – there is so much we can do with data and AI, that it’s very easy to quickly create a long list of ideas. Now that’s a great start, as we can then review them in terms of business benefits (eg time-savers, money-makers, USP-creators) versus the estimated effort to deliver them (eg quick/simple, hard/complex). We can then prioritise them with the quickest, simplest and highest value use cases first. So, what are your priority use cases?
  3. Our data access – in order to harness AI and ChatGPT-style technology, we need to point them at data. To state the obvious, we need the data and information available in the first place, in order to apply the technology, in order to deliver value. We can’t ask questions about data that doesn’t exist or isn’t available. Think about both internal & external data sources here – from financial and customer data through to public data or, my favourite, weather data. We need to ensure that the data and information we want to interrogate are accessible and available. So, is your data available?
  4. Our data quality – in order to derive quality insight from our data, we need good quality data in the first place. Garbage in = garbage out, and all that. For many use cases, our data will need to be accurate, ordered and structured. For some use cases, unstructured data could present the opportunities for automation. So, is your data at the right quality level for your priority use cases?
  5. Our governance & controls – every organisation needs to be focused on data security and compliance. We need to control data access both externally and internally. We need to ensure we’re aligned with GDPR guidance and the evolving AI guidance (see our December issue for more information on this). We need rules, policies and systems to do this. So, are your data governance & controls in place and updated for AI?

This is a lot to think about it, isn’t? It might feel a bit overwhelming or unwieldly or cumbersome or complex.

So, how do we fast-track innovation & get started ASAP?

I’m a big fan of time-boxed prototypes to prove a concept quickly, cheaply yet effectively. Over the years, I’ve successfully used this approach when I’ve worked within organisations, and also when I’ve been consulting for organisations. By successfully, I mean that it either proves a concept or it doesn’t!

The approach I take is:

  1. Define a realistic timeline to prove the value – for instance, 6 weeks is often a decent amount of time to test something.
  2. Define the minimum resource required to prove the value – ideally this would be kept to a minimum and 2-3 people, with one person to think commercially and 1-2 people to build the tech solution.
  3. Define the use case – choose a specific, scope-restricted use case that will be possible to deliver in the timeline and will get people excited about the opportunity…perhaps something topical or high profile within your organisation?
  4. Start with 1-2 data sources – select the minimum amount of data you need to prove the value. This might be in terms of the number of data sources, ideally starting with one. Or this might be in terms of the time period, for instance, just using the last X years’ worth of data based on what’s easiest and quickest to access. Start small first…you can always expand it once the value is proven.
  5. Run this in the opposite way to your usual projects so it feels completely different – for instance, get the team in one room in a different venue for the project duration, map everything up on the walls, call the project a codename, create a project culture where everyone eats together…whatever it takes to make it feel different
  6. Present the solution in a different way – for instance, specifically target the stakeholders you want to influence, create a video demo of your solution, display outputs on the walls, walk them around a room to explain different parts of the solution…again, whatever it takes to make it feel different.

Your goal is to prove the value to your organisation, excite people with the opportunity and demonstrate what can be done in X weeks. To win the trust, confidence and support to do more projects like this in future. So, don’t just do what you’ve always done and expect different results.

Are you ready to build your first ChatGPT-style data interrogator?

Well that is exciting! For both our business, and for our clients, we are harnessing the functionality of ChatGPT and AI…but locked down to an organisation’s private data. This enables the team to ask a question to interrogate their own data, in a secure, confidential way…and get an answer based solely on their organisation’s data.

Here are a few examples…

We’re exploring three types of use case:

  1. Ask questions about content within unstructured documents eg Word docs, pdfs – for instance, you might want to say “summarise this document for me” and receive a description of the content of the document = the benefit could be to save huge amounts of manual effort for your team
  2. Ask questions about data within a database without using code – for instance, you might want to ask “what is the average profit by client in 2023?” or similar and get an immediate answer = the benefit could be faster access to business insight to inform quicker decision-making
  3. Train AI to do a specific task to help your customers – for instance, you might want to create a querying tool or a summarizing tool for your clients to use based on your data = the benefit could be a new way to create a commercial advantage and USP by monetising your data

Think about your organisation – what type of use case resonates most with you when considering your challenges and opportunities?

To get started, I like to approach this in 5 steps:

  1. DESIGN – design the solution based on the priority use case, including selecting the right tools, developing the prompts or queries you want to answer & creating the plan, timeline, resources & costs
  2. CONNECT – connect your selected data sources after considering if the data needs cleansing or preparing first
  3. SETUP – configure your private AI tool so it’s secure and private, such as Microsoft’s OpenAI
  4. PLAY – experiment with prompts and queries to answer your priority questions with an experimental research & development-style mindset…be prepared to play!
  5. LAUNCH – deploy your solution into production and train your team in how to get the best value out of it

Want some help? Drop me a line and we’ll chat about how we can fast-track your AI roadmap.

Thank you for reading…I’ll be back soon with more data news.

Remember…if you’re in data hell, I’m data Hels!

Recommend a friend or colleague to join our mailing list for DATA DATA DATA –hello@data-cubed.co.uk or

Connect with me on LinkedIn – https://www.linkedin.com/in/helentanner/

Email me at helen@data-cubed.co.uk

This is me and the Bristol-based team at our team day in January…and, yep, they really are that smiley!