Cut the AI Noise – Show Me the Benefits
One question that we often get asked from business and technical leadership in enterprises of all sizes is, “with so much noise around AI,
how do I demonstrate sustainable and quantifiable business benefits that scale reliably”.
In 2023, we saw the democratization of access to AI. A recent PWC survey shows that almost every business leader is saying that AI will be critical to their organization’s success over the next five years. Among Fortune 100 companies, 15% have a designated leader for firmwide AI usage strategy.
Engineering leaders are among the most interested and engaged audience in AI usage. Though they are excited about the productivity, speed to market, and cost efficiency gains possible with AI, questions such as:
- My engineers are using AI generated code, but who is making sure it is bug free?
- Are we infringing anyone’s IP by using AI generated output?
- Is our enterprise and customer data safe when using public Large Language Models?
- What would be the cost of using AI at scale?
Are still on their minds, and any solution delivering AI benefits needs to address such concerns.
We have outlined a simple framework below that can help provide clarity in deploying an AI solution in enterprises.
Define, Refine, Get Going
The first step to benefitting from AI is identifying functions in your organization that have persistently been a bottleneck in your efficiency. Speed to market can be a challenge to address, however in an enterprise setup multiple factors contribute towards go-to-market. Identifying a factor that’s often a troublesome spot is where you can look to AI for efficiency gains.
The second step is evaluating the inherent risk of a function when introducing AI to it. Selecting a low-risk function has several advantages, the most prominent being protecting business continuity.
Lastly, you should be able to quantify benefits either as top line growth or expense reductions that can be directly attributed to the AI solution. Go for tangible outcomes vs outcomes that require further work to translate into benefits.
Example of an AI Implementation Journey
Watch out for Key Considerations Along the Way
As you proceed to implement an AI solution for a specific problem, it’s important to consider the time necessary to implement it, which depends on the solution’s readiness. Though many open source or vendor supported AI models are available, not every model is trained in the problem area you are trying to address. Training a model can take time and resources, which can increase the total cost. The accuracy of a model’s outcomes depends on how well it is trained. You may need to look for AI models that are already trained and have a high enough accuracy for solutions in the problem area you are considering. Because AI is inherently probabilistic, you shouldn’t set out with the expectation that it will guarantee an accurate solution for every scenario. There may still need to be vetting done to de-risk any unintended outcomes.
Share the Learnings and Be the Champion
The power of AI gets multiplied when it learns from a diverse set of resources in an ecosystem. In an enterprise setting, it means scaling the use of AI solutions horizontally when you feel confident about outcomes in one area. For example, if you are utilizing AI for product use case design within one team and are happy with the results, then it might be time to expand that usage to other teams on the same product and then in teams on a different product. This can help the solution learn across use cases and products and provide more contextualized outcomes.
At Saralam, we are witnessing the power of network learning in our No Code Testing service, where new applications benefit from test cases discovered across a vast existing application base.
References
[1] PWC survey
