T-Shaped #1: The dawn of open source AI and what does it mean for you and (small/medium) business?
Welcome to the first edition of the T-Shaped Newsletter!
May 30, 2023
Welcome to the first edition of the T-Shaped Newsletter. I just had to kick off this publication with the topic that should be of utmost interest to all T-Shaped out there: how to build real business value with AI. I believe that generalists with selected domain expertise are destined to lead businesses through the current AI revolution, and what they need is a good framework to do that. This will be the topic of our first edition.
My recommendations are based on our own internal work at our software company and ad agency as well as an AI-tech-based project that we did for our clients. We’ve been doing natural language processing solutions for over four years now, and within the last few months, we did a deep dive into GPT. Here are our learnings.
The main problem with AI for business
There are hundreds of How to’s, 15 Ways to… and Twitter threads showcasing different prompts and ways to use new AI tools that are popping up every day.
This deluge of new solutions is as fascinating to specialists in given fields (copywriters, SEO specialists,. etc.) as it is debilitating to leaders and people responsible for the strategy and value of the companies.
Here I propose a framework for companies to experiment and develop their own solutions based on the AI models: Rapid AI Development Sprint (RAIDS).
Small and medium business
I’m not talking here about airlines, car manufacturers or big pharma, but about small and medium businesses, consultancies, ad agencies, eCommerce companies, service businesses, and manufacturing companies. Almost every business that has the right team and growth mindset can benefit.
Why now? What has changed?
Until recently, the approach to the current AI revolution was doom and gloom for smaller companies because AI was about to be monopolized by a few big players. Right now, the tide in AI seems to favor open-source solutions and smaller teams working on specific implementation.
As leaked, Google documents say:
While our models still hold a slight edge in terms of quality, the gap is closing astonishingly quickly. Open-source models are faster, more customizable, more private, and pound-for-pound more capable. They are doing things with $100 and 13B params that we struggle with at $10M and 540B. And they are doing so in weeks, not months.
Read great blog post by Simon Willison.
While Open-source alternatives to players like OpenAI are proliferating, they are preparing to release their new open-source language model. Pressure may grow on Google to release more open-source AI, too.
Examples like Midjourney, which was 11-people strong, and built a much more robust image generation solution than Dall-e, are proof that small, focused teams can surpass big tech in AI fields.
I’m not saying that you will build the next GPT, but opportunities for innovators in the markets are opening. With the right approach and small investment, you can hugely increase the value of your business with AI. OK, how to approach this problem? Meet RAIDS.
Rapid AI Development Sprint (RAIDS)
Where to start?
Stop looking for the perfect AI tool stack and start thinking about the needs of the businesses. You can perform RAIDS in two to three weeks and get tangible results that will significantly increase the value of your company and future-proof it. There are three stages of RAIDS:
Stage 1: Discovery and Analysis
Map your business processes and find high-impact areas for intervention. Think about data that you own and how you can use it to train your specific model. Select available tools, frameworks and technologies for the project and develop the roadmap.
Map of opportunities
Map the company's areas that present some automation or improvement opportunities. You can use good old and tested business model canvas to help you look into different parts of your business and look for opportunities to automate. [Right now, I’m working on a specialized RAIDS canvas developed based on my experiences working with companies. Stay tuned.]
Stage 2: Design and Prototyping
Start iterating on the project with your development team. Prototype fast and try to provide updated models/products daily to get your company's feedback.
Data = gold
What kind of data do you own? Your company's proprietary database might become one of your most valuable assets. Look into your cloud storage, project management systems, CRMs, marketing automation suites, blogs, manuals, product descriptions, service delivery processes manuals, etc. Search for anything that can become training data for your AI solution and is specific to your company and the ways it delivers value to your clients.
Stage 3: Proof of Concept and Value Demonstration
Integrate all the elements of the prototype you’ve built into one solution and provide an internal demo for the company. If possible, share it with external stakeholders (e.g. clients, investors, your mom). Sum up learnings and discuss outcomes of the RAIDS and decide if the opportunity for the company is worthy of further pursuit.
A real-life example of RAIDS
We worked with a manufacturing company that had a comprehensive database of product information and associated services. The challenge was to make this vast amount of data easily accessible and usable for the sales team.
Our solution was an AI-powered internal chatbot developed during RAIDS. This tool enabled sales reps to quickly and accurately find product information and related services, reducing the need for lengthy consultations with product specialists.
The chatbot became an efficient assistant, helping the sales team craft more personalized offers by learning from past implementations. The benefits were: quicker access to solutions, less consultation time with senior staff, and better matching of offers to customer needs.
Team composition
You’ve guessed it: optimally, the Team for this kind of project should be led by a T-shaped person with a broad understanding of your business and particular knowledge in designing and leading product development.
The optimal candidate would be a product manager with some software development background and AI solutions experience. I know it’s tough to find this kind of person, but at least this should be your North Star. If in doubt, optimize for high intelligence and need for learning and curiosity.
The RAIDS team should have good programming talent that can cover both front-end and back-end tasks. There will be a need for people experienced with working with different kinds of APIs and databases. Once again: if in doubt, optimize for fast learners in the team.
You will also need someone that will help to test the solutions and provide quality assurance and someone that can think in user experience terms. BTW - a good T-shaped leader of the team will take care of UX.
To sum up: your RAIDS team will probably consist of a T-shaped product manager, 1-3 developers, and a tester. Not bad for a squad that will deliver on the future of your business, don’t you think?
Tips and things to remember
Make it special; make it an adventure.
You can communicate this process as a kind of hackathon - something unique and worth celebrating in the company. You may put some extra effort into that, but it can serve as a great break from the business's daily chores for the people who feel engaged in the company's future.
🤗 - remember this emoji
Hugging Face is a key resource for any AI builder. Its Transformers library, Model Hub, and Inference API offer easy access to state-of-the-art pre-trained models and infrastructure, reducing the barrier to entry in using advanced NLP technologies.
Progress every day
For the success of this kind of project, it is crucial to provide daily prototypes and keep up the momentum. AI-based solutions tend to follow the critical mass mental model - they don’t show any value until they do.
There is no linear progression that can be easily observed, so you have to iterate for a predetermined amount of time and tweak the models and approach every day of the sprint.
Last but not least: Beware of thinkism
“There’s something I call thinkism, which is this reliance on trying to solve problems by thinking about them.”
— Kevin Kelly
I love KK, and you should too. His wisdom is also applicable to AI solutions development. There is no amount of thinking, planning, and reading that you can do that will lead you to an outcome.
It’s an iterative, experimental process, and you’re building, gathering data, and applying new learnings every day. Only this approach will lead you and your company to tangible results.
PS.
This is the first edition of T-Shaped, and your feedback is the most precious thing we can ask for. Drop me a line if you have any comments, and if you like what I write, you can share it with friends or colleagues.
What's next? T-Shaped aims to connect different dots: Business models, building and managing top-of-the-class teams, growth hacking, product design, dark patterns, and psychology of consumer and user. The next edition will be about buyer personas and how to research them and use in your marketing and product design.
Contact us today,
start working tomorrow
Oops! Something went wrong while submitting the form.
