Bret Taylor, Chairman of OpenAI, recently expressed the belief that software programming will be eliminated as a job in the future. We’ve also begun to see workforce reductions take place across some industries in the name of AI automation.
Does this mean we don’t need humans to run our businesses anymore? Not at all, and I think this way of thinking only limits the potential of what we, as humans, can accomplish with AI. But it does beg the question: What does this say about the future of services companies that have traditionally been the software implementation arm of businesses?
AI enablement is much broader than automated code writing. Indeed, AI models are getting more intelligent and cheaper, and increasing amounts of code are being written by AI. At the same time, as humans, we are moving to become designers of agents and multi-agent systems augmenting our businesses. Programming tasks are becoming agentic orchestration tasks, and rather than writing deterministic code, we are designing semi-autonomous agentic systems tailored to our business needs.
This calls for a whole new approach to technology. How do we set up and manage a team of intelligent, non-deterministic, semi-autonomous artificial agents?
1. PRIORITIZE SPECIFIC DATA
My company, Cognizant, is on its own journey of agentification, and we have learned that this process does not happen automatically, nor should it be left to happen organically. Thinking that a single “AGI” agent will be able to run our businesses out of the box is rather naive. The future is multi-agentic.
Agents in an agentic enterprise need to be grounded in the specific data that flows through the business. AI agents will have to operate robust API and microservices layers. Uncertainty and nondeterministic behavior need to be accounted for and mitigated, along with all other responsible AI practices and safeguards. This all takes oversight and governance.
2. SET UP FOR INCREMENTAL EXPANSION
An agentic enterprise needs to be set up for incremental expansion. This should allow the introduction of new agents from different sources within the company and sometimes from third parties, so interoperability is an important requirement—and this necessitates complex oversight.
For example, the introduction of new agentic sub-systems will have to go through a sandboxing and validation process tailored to the business, and the modularity and granularity of the agentic components need to be carefully designed and maintained. The cost/benefit/security analysis should inform the choice of LLM and the manner of hosting for the agent, and future-proof extensible inter-agent coordination mechanisms and architectures are also required. This nuanced approach to kicking the tires—through testing, design, and performance analysis—can only effectively be managed by humans.
Recent work in our Cognizant Moment digital experience team has highlighted the fact that, quite often, the user experience for human/agent interfaces will have to be designed depending on the context and user persona. The UX must also be based on a human-centric, entry-point agnostic basis, and be exposed in a unified manner through a diversity of channels.
3. MAINTAIN AND IMPROVE
Finally, an agentic enterprise is not a fire-and-forget proposition. It requires maintenance and continuous improvement. Logs from each agent in the multi-agentic fabric of the business need to be regularly monitored and mined for improvements and role-specific fine-tuning. Customized dashboards for metering and various context-specific technical and business measurements should be provisioned.
In our AI research labs, we are taking this even further by innovating on dynamic agentic architectures; leveraging machine learning to reconfigure agentic workflows, improve resource allocation, and imbue agents with continuous learning; and automatically adapting agentic systems to new tasks by automated creation/merger of agents.
As a leader in the field engaged in agentifying our own business, as well as that of our clients, we see a lot of work ahead of us. We also see an important role for humans in this multi-agentic future. IT services companies will always be critical to the last mile of infrastructure modernization for enterprises, and human ingenuity and critical thinking play a key role in how this takes place.
As we look ahead to the future, AI agents will continue to drive even greater efficiencies within businesses by augmenting human capabilities. While some reskilling will be necessary, the irreplaceable value of human oversight, creativity, and decision-making remains essential in this transformative journey.