Applied Generative AI
There are many definitions of AI, and many AI technologies with differing capabilities and uses. Generative AI is composed of multiple technologies as well. Generative AI is moving us closer to what is called a “sociotechnical society,” in which people and machines are more integrated. Most of the recent public conversation and media attention has focused on chatbots and their accuracy, hallucinations and potential for societal impact. Generative AI changes the way innovation is organized and realized — especially the path from concept to applied innovation and accelerates the business impact. It is already convincing in the solution of niche problems, but it still lacks creativity and business strategy. The innovative value of Generative AI based on a large language model (LLM) such as GPT appears to be incremental in nature. Other advanced generative algorithms used in CAD, for example, are more mature, and they are built for problem-specific human-machine collaboration. They take into account many solution parameters such as performance or space requirements, materials, manufacturing methods and cost constraints.
Innovation through Coworking AI - Centaurs
Innovation and R&D teams that are able to facilitate Generative AI capabilities in their process of disruption are likely to be the better innovators. Centaurs — teams that are half-human and half-AI — coworking in design, development and engineering outperform human-only teams in speed and exploration of the solution space. For centaur teams, human knowledge must be extracted, structured and formalized for use by AI machines. Structured data can be translated into knowledge with AI support, closing a loop of knowledge utilization. This quickly weeds out hypotheses that do not fit into the solution space.This enables smaller development teams, in particular, to catch up with those of the industry giants, at least in coming up with better and lower-cost products and solutions.
Develop Minimum Marketable Products
Generative AI is giving human innovators superpowers by maximizing the “What-ifs per hour” and increasing the rate and number of iterations. Organizations are able to move from the classic Minimum Viable Product (MVP) toward a Minimum Marketable Product (MMP). In the past, innovation teams often aimed for an MVP to prove their concept, value or targeted technology. With the advent of Generative AI, these MVPs are evolving into MMPs. The creation of rapid iterations for user evaluation is being revolutionized by AI products that enable detailed, professional-looking mock-ups. Testing with live customers, e.g., through a fake-door test, leads to direct development insights. Those tests reveal:
Product target audience
Willingness to pay
Product positioning
Messaging and go-to market strategy
The Right Prompt for chat-based Interactions
Some conclude that ChatGPT, Bloom, Bard, Aleph Alpha, Bedrock and other LLM tools will revolutionize idea generation by asking the right questions. These tools can generate input that reflects what they have been trained with, but they are unable to come with fully original ideas. Those tools are not capable of critical thinking and gain no deeper understanding. Nonetheless, these chat-based interactions have particular value in producing ultra-fast rough sketches for:
Market sizing and segmentation
Product and competitor landscapes
Competitive advantages
Summarized customer expectations
Assumption mappings
Customer journeys
Value proposition creation
Generative AI is Changing Operations and Productivity
Generative AI also represents a new frontier when it comes to business operations and productivity. The short-term use cases will primarily be around product development, customer experience and interaction, and employee productivity.
Absci — a drug creation pharma company — has launched a generative AI tool to assist with the identification and testing of antibodies typically not found in existing databases. It anticipates this advancement could cut the time to clinical testing by more than 50%.
Autodesk — a software firm specializing in design — uses its Dreamcatcher platform to generate CAD designs based on user goals and constraints. The software can incorporate elements such as local regulations and sustainability into its various proposals.
IBM and the Masters — a U.S. golf tournament — recently used generative AI to provide spoken commentary to highlight clips for the 2023 tournament. This will result in more than 20,000 highlight reels having audio narration.
SKT — a South Korean telecom provider — has dramatically expanded its AI capabilities and services. During the pandemic, it created Nugu Care Call, an AI-powered check-in call service to assess for symptoms of COVID-19 and assist healthcare workers.
Stitch Fix — a clothing recommendation service — has incorporated AI to assist in writing its advertising headlines and product descriptions. For copywriters, this means significant time savings.
Accenture’s velocity team — has been using Amazon CodeWhisperer to help complete code based on natural language prompts. As a result of this partnership, they’ve seen a 30% reduction in development efforts.
Synthesize Data for Faster AI Development
Innovations get more and more data-driven, and many ideas demand large datasets with partly sensitive information about Customers, Suppliers, Transactions, Financials. Valuable operational data and Confidential corporate data. Training an ML system can require datasets that exceed the available number of samples. Training with insufficient and sparse datasets reduces the chance of success.
Generative AI solutions are able to synthesize datasets for training purposes. Some of the solutions work with table-based information, and others provide simulation input, for example as generated audiovisual and 3D assets to simulate sensory input. Synthetic sample datasets of key enterprise databases can provide training material with a controllable level of bias, for it will speed up general AI adoption within the organization without creating operational risks.
The beauty of coworking/innovating with AI in general is that, after successful prototyping and hypotheses evaluation, the core components for value generation are already built and the model is trained. There still is work to be done to scale the model and realize revenue, but this is not uncommon for innovation units. The difference in working with an AI is that the developer needs to be clear about ethical and legal concerns, data privacy, possible malicious data imports that drive bias, and the ability to explain the models and products developed.