The Right Shift?

Shift Left

No piece of software is completely defect-free. Companies try and eliminate as many bugs as they can before they release the products to customers. However, to meet the release dates, products do ship with known and unknown bugs. The known ones get taken care of in subsequent patches and releases, and the unknown ones cause blue screens etc, and are reported by customers. The later the bug is caught, the more severe is the impact in terms of cost to correct and customer experience.

Companies therefore try and catch bugs as early as possible in the software lifecycle. This is the ‘shift-left’ principle. If the timeline stretches from left (where the product starts getting conceptualized) to right (where it is released to the customer), then the aim is to shift the catching of the bugs from later (from the right) to earlier (towards the ‘left’). A similar logic applies to incident management. Instead of reactively logging incidents, the effort is now to proactively place alerts towards the start of the business process (the ‘left’ side), rather than know towards the end (the right side) that the process has failed. These alerts help to take corrective action and prevent incidents from occurring in the first place.

The GenAI impact

However, with GenAI, this principle has been given the short-shrift. The latest technologies and advancements are being released to the market, and the burden of testing has been transferred from the company to the customer. The accuracy of the results is in question, with models hallucinating as indicated by some of the responses. The closed nature of these models and algos only accentuate the problem. This shift ‘right’ has even led to hasty product withdrawals due to customer backlash. OpenAI dissolved the team focusing on product safety following the departure of its chief scientist, and later created an oversight board committee to evaluate safety of AI models. These incidents will reinforce the ‘buyer beware’ mentality in businesses. And that is not good for the GenAI providers: slower adoption leads to lower ROI. While the stock prices of GenAI providers have shot up, revenue is to catch up.

A different view

While it is early days to state either way definitely, the safety and transparency approach has not been abandoned completely. Two companies seem to be taking a different approach to GenAI and LLMs. And the role of IT service providers may be key to adoption.

Meta has released an open-source LLM Llama. The code base can be modified, and the models installed in a secure environment enabling enterprise data security.

Apple announced its Apple Intelligence suite that will supposedly bring the control and privacy back to the customer with its on-device GenAI features, while enabling it to draw on personal context for providing relevant information. While a lot will depend on how the implementation pans out, this helps put the responsibility back on the provider for rolling out products that put the customer at center. While Apple is not the first mover in this space, it has the scale of 1.5 billion userbase to effect a shift in what to prioritize in the GenAI race.

The role of IT services companies will also become important. They’ve played an important role in package implementations in the past, where the help with customizations and help resolve critical bugs with the product companies before the implement the package for the (enterprise) customer. A similar role will become important when a customer wants to implement GenAI solutions from various providers. Not all enterprises will want to be early adopters of the GenAI providers’ latest releases. The services companies can help evaluate the suitability of the product for the customer’s use cases, make sure that the performance is as per the customer’s policies (or with an acceptable level of deviation), can help bundle various provider’s offerings into an integrated solution before deploying these for enterprise customers. (Read also The Operating Model for GenAI | LinkedIn)

Conclusion

The providers will always want to push out the latest features, to differentiate themselves from competitors and to gain market dominance. Businesses will want to make sure that the adoption is non-disruptive to end customers. The biggest players are putting stakes in the ground in this disruptive battlefield based on their evaluation of the market dynamics and their respective strengths and weaknesses, and hopefully many different positions will exist and be successful.