Our learnings from working with generative AI model ChatGPT
- Posted on February 13, 2023
- Estimated reading time 5 minutes
I’ve had a long career in tech, and I firmly believe that in the AI-first future we will look back at the democratization of AI through Large Language Models (LLM) like GPT as a true inflection point in our industry. The slew of OpenAI integrations recently announced by Microsoft will be the catalyst to drive AI’s global ubiquity, access and adoption. It may feel like hype right now, but what we’re seeing with ChatGPT/generative AI is monumental for businesses and society.
Before I go on, a quick primer: generative AI or, generative artificial intelligence, is a type of machine learning that can cognitively generate synthetic data (like blog posts, reports, program code, artwork, videos, etc.) rather than simply analyzing or acting on existing data. GPT is the underlying LLM model and ChatGPT – announced by OpenAI in November 2022 – enables a conversational interaction with the user.
Avanade believes that AI is the next wave of computing. Like mainframe to client server, evolving to mobile first and then to cloud first, our clients need to be thinking, operating, and moving to AI-first. Indeed, I believe this is the year that AI will go mainstream. In the same way that Mosaic and Netscape Navigator democratized the internet, the accessibility and usability of ChatGPT brings AI into our everyday lives.
What we’ve learned experimenting with generative AI
Since Microsoft first launched a series of services in preview with OpenAI back in early 2021, Avanade has been at the forefront of experimenting and collaborating with Microsoft on GPT. We have invested significantly in our responsible technology approach and how generative AI needs to be used responsibly. We were early with Github Copilot and Codex, an AI programmer assistant and working on our plans to scale up what we saw in our own tests as significant productivity gains. And we have worked on how we embed DALLE-2 capabilities responsibly (image generation) in delivering business results. This early access has provided us with experience and practical application of the technology to help organizations maximize value and solve real-world business problems.
How organizations are using generative AI today
We are actively working with clients across the world who are harnessing generative AI. A few examples:
- A service company is evaluating the performance of generative AI models compared to established standards for customer document validation (e.g., compliance, completeness and accuracy) for tasks like intent classification, key extraction and text normalization.
- A manufacturer has identified the opportunities to use generative AI to inspire designers, engineers and marketing teams to create and deliver new product designs.
- A non-profit organization is exploring the use of generative AI to assist as a copilot to generate grant reporting, redirecting time spent on related administrative tasks toward delivery of the non-profit’s core mission.
- An oil and gas company is improving the knowledge management and search results on trouble tickets by integrating generative AI and the existing enterprise corpus and ontology.
This is just the beginning, and it will only become easier to create better ways of working that increase efficiencies and free employees to focus on high-value tasks and personalized experiences. For example, imagine a future where a financial advisor can meet with her client using simple text prompts, generating a Teams personalized meeting in a custom virtual world with personalized information, charts, and the advisor guide for the discussion. Expect generative AI applications and use cases like this to become prevalent in our everyday lives over the coming 12-18 months.
Gartner predicts that by 2025, generative AI will be producing 10 percent of all data (currently it is less than 1 percent) with 20 percent of all test data for consumer-facing use cases.
By 2027, 30 percent of manufacturers will use generative AI to enhance their product development effectiveness.
Keep trust at the core
Even as the current GPT model (currently v4.0) is scaled, and the accuracy and relevancy of artificial intelligence improves, a human ‘copilot’ is essential. As Satya Nadella said, “AI is about alignment with human preferences and societal norms.” Creating and maintaining trust must be at the core of anything involving AI. Avanade has developed a responsible AI framework to help our clients align the adoption of technology with their company values and behaviors. Based on our work with these clients and our involvement in digital ethics policy development groups, here are five tips to responsibly innovate with AI, including GPT and ChatGPT:
1. Set the right foundation: Start simple in areas where you may have already deployed capability, like an existing chatbot. Select the use cases where introduction and integration of OpenAI capabilities can be done in a modular and scalable way with continuous evaluation of the model’s performance.
2. Prepare your people: Generative AI is not all about cost cutting and automation. Change enablement will help users work iteratively from generated concepts that need to be tweaked, refined, enriched and approved. AI is the productive assistant, not the replacement.
3. Appoint responsible AI ambassadors: Select for specific use cases and consider the availability of quality, diverse data for the domain, and the ability to incorporate human subject matter expertise feedback (AI Ambassadors) for rapid, iterative re-training of the model.
4. Establish governance: Consider a centralized function to investigate, validate, refine and channel AI across your business. There needs to be a robust approach of transparency to and communication with customers, partners and employees on risks, limitations and uncertainties. Risk control and mitigation plans need to be aligned with the affected stakeholders and legal requirements.
5. The most important question – why?: Generative AI, as with all other technology trends, will only bring value if you truly understand why you are using it. What problem are you solving? Does the business case stack up? What outcomes and benefits do you really want to achieve? Being clear as to why this is good for your business and not just a thing to implement is fundamental to embracing generative AI.
Get started with generative AI now
Avanade has been working with OpenAI since its early integration with Microsoft. The AI models will become even smarter and more flexible and so organizations need to lay the groundwork and start testing now to leverage the best of generative AI in a responsible way.
I invite you to contact us to learn more about our Responsible AI framework and opportunities to drive business value from AI.