Gen AI: From theory to practice
We know that hands-on lessons from real-world scenarios are invaluable, especially as technology is changing so fast. Here are our recommendations:
1. Identify use case viability
Generative AI presents a wealth of opportunities for enterprise leaders across industries. They can harness data to generate insights that support new ways of working, innovation, and the ability to make decisions at speed and scale.
While many use cases exist, generative AI is not a one-size-fits-all solution. The first step is to understand your business' critical challenges where it can deliver the greatest impact.
For example, one of our clients was spending too long writing reports. It now uses generative AI to create financial intelligence reports from vast volumes of financial data, saving significant time for hundreds of stakeholders. Another client needed to boost sales. With generative AI, natural language processing, and other AI and machine learning capabilities, it can analyze customers' chat interactions with sales representatives to quickly identify upselling and cross-selling opportunities.
2. Develop talent and expertise
Once you've identified promising use cases, ask yourself – does your business have the resources to bring them to life?
We suggest creating a generative AI center of excellence (CoE) to train employees for new roles like prompt engineers, prompt-compliance checkers, and customer-protection officers. A CoE also helps analyze infrastructure, process, and budget requirements to give you a clear view of the potential costs and benefits.
By channeling projects through a CoE, you can effectively democratize ideas, allowing everyone to bring generative AI to their area of work, while simultaneously limiting production to the most promising use cases.
Remember, organizational maturity can increase if you invest in upskilling employees. At Genpact, we've developed a learning platform, Genome, to keep our 115,000+ employees up to date with emerging skills, including generative AI.
3. Build trust with governance
While generative AI presents endless opportunities, it also creates new responsibilities.
Data fuels generative AI, and an enterprise's data is often its most prized asset. Therefore, bolstering data security through your governance policy should be a priority. There have been many reports of people accidentally leaking sensitive company data through public generative AI programs, causing irreparable damage.
Your governance policies should also encompass auditing mechanisms and tackle legal considerations, as gray areas abound. Already, there are court cases around intellectual property rights and large language models, and governments worldwide are penning new legislation. With so much changing, working closely with your legal team is essential.
Finally, you must include a plan for the ethical use of AI with the right guardrails, prototype delivery system, operating rhythms, use cases, change management, and more. And if that sounds overwhelming, don't worry; you can use our framework for responsible AI. Because with trust comes greater scalability.
4. Educate stakeholders
So, you've chosen a pilot project, have a team in place, and incorporated responsible AI safety measures, and you are ready to start. At this stage, educate stakeholders on what they should and shouldn't expect from generative AI.
For instance, the technology still struggles with hallucinations – giving convincing but false answers to questions. In the early stages, expect a good deal of involvement from subject matter experts as they train applications to increase accuracy and usability. And factor in the cost of multiple rounds of prompt tuning and additional labeling based on their feedback.
Similarly, depending on the model you use for generative AI, the amount of data you can process may be capped because of token limitations, hampering the speed of innovation. You must also ensure development and cloud teams work together to optimize resource allocation and mitigate potential challenges.
Lastly, make sure your stakeholders have visibility throughout.
5. Tap into your partner ecosystem
Partnerships can prove invaluable if you still have talent gaps. Whether you need consulting support to identify use cases or technical expertise to design a custom large language model, partners can augment your talent pool.
They can be especially valuable for generative AI programs. Whereas in-house talent may have only worked on a handful of AI projects, technology service providers can share their experiences from leading hundreds of projects across industries.