Skip to main content Skip to footer

January 19, 2024

A practical guide to introducing gen AI into the enterprise

Generative AI isn’t like other technology rollouts. Here are 10 things companies should do to securely and effectively deploy this powerful technology.

For the past year, generative AI has been positioned as a technological breakthrough for enterprises. And yet, despite an overwhelming consensus on the technology’s value, many companies are taking a cautious approach, limiting or even restricting its use.

But is sitting on the sidelines playing it safe—or taking the biggest risk of all: falling behind?

The reality is, the value of generative AI—up to $1 trillion in annual productivity gains by 2032, as forecasted in our recent study—far outweighs the perceived risk, as long as the technology is introduced in a thoughtful way. Companies shouldn’t outright restrict generative AI, so much as define the boundaries of appropriate use that protect the organization from any number of issues that can arise, from runaway costs, security and compliance to workforce engagement.

Here are 10 ways companies should adapt their traditional technology roll-out processes to harness the power of generative AI in a safe, secure and effective way.

10 ways to succeed with gen AI rollouts 

1.    Take a “light” approach to business case development

Because generative AI is a novel technology, many aspects of the standard approach to business case development do not apply. For example, companies don’t need to spend time and resources establishing the relevance and value of the technology, since that is widely accepted.

They can also skip other aspects of the business case, such as quantifying cost savings from specific use cases or identifying which parts of the tech stack or which processes generative AI would replace. There is simply not enough information or understanding at present to define those areas.

What companies are left with is a “lightweight” business case that focuses on the value of experimentation and learning, rapidly identifying where and how this technology can be used effectively, as well as the boundaries needed to ensure safety, security and responsible use.


  • Fast-track pilots, POCs and other programs to build knowledge, skills and capabilities. 
  • Leverage experimental funding or seed funding to rapidly test and learn.


  • Delay getting started or wait for clarity.
  • Rely on lengthy and complex business case development processes. 

2.    Start with simple use cases

As my colleague Surya Gummadi, EVP and President, Cognizant Americas, wrote in a recent post, “Before looking at [use cases that drive] radical gains, businesses should start by unlocking value trapped within their existing business processes.”

In other words, companies should select relatively simple use cases, even though generative AI can do some fairly advanced things.

Examples include creating a knowledge repository for call center agents or automating routine developer tasks. These are good places to start because they are internally focused.

On the other hand, customer-facing applications may open the organization to more risk since the company will have less control over how people interact with the content produced by the model.


  • Focus on internal use cases that will build a foundation for future applications.
  • Ensure data readiness to tailor the solution to business and customer needs. 


  • Launch customer-facing applications as an initial deployment.

    Don’t overdo it on technology evaluation

As I’ve previously said, most generative AI tools available today offer comparable capabilities. Because there’s little differentiation, companies should not spend significant resources on evaluating the solutions offered by hyperscalers or other leading tech organizations. Such analysis is unlikely to reveal significant differences between the tools and is also likely to be quickly outdated. 

Instead, turn to existing partnerships to launch preliminary use cases. When selecting a new partner, consider non-technical criteria like trust, integration and experimentation. 

When working with niche generative AI players, and startups in particular, a more in-depth evaluation may be necessary to ensure the tech provider offers a level of security, privacy and compliance often accepted as standard among hyperscalers.


  • Consider working with hyperscaler partners to launch preliminary use cases.
  • Focus on trust, integration and the ability to experiment, in addition to technical capabilities, when selecting a partner.


  • Conduct an elaborate analysis of technology providers’ capabilities. 
  • Place trust in niche players or startups without a robust evaluation.

4.    Invest in external expertise

Out of the eight billion people on the planet, less than 200,000 know how to operate an AI system, and a mere 50,000 can explain how generative AI tools like ChatGPT work, according to Dr. Vishal Sikka, an early adviser to OpenAI and CEO of Vianai Systems. This begs the question: Does your company have the internal expertise to move the program forward?

Unfortunately, the odds are not in your favor. Most organizations will need to tap a transformation or technology partner to help evaluate platforms, introduce it to the organization and identify the other technology elements needed to integrate it across the enterprise.


  • Assess your organization’s resources and where you need external support.
  • Get help bringing the technology into the enterprise.


  • “Go it alone,” even when executing relatively simple use cases.
  • Form strict or exclusive partnerships with vendors that will limit choice in the future.

5.    Design an adaptable system architecture

Companies will need to develop a net-new, end-to-end architecture and strategy to support the use of generative AI. This includes identifying all critical program components, such as integrations, data sourcing practices and maintenance. 

As time goes on, the system will need to continuously change as the organization learns, the technology matures, use cases evolve and the regulatory landscape becomes clear. Companies need a flexible framework that can change with the business’s needs, goals and capabilities, as well as adapt to market and regulatory forces.


  • Think creatively and progressively when designing the framework.
  • Understand that the architecture will need to be adapted over time.


  • Create a rigid end-to-end system based on how the technology works today. 
  • Rely on existing processes to support technology as it advances.

6.    Establish new security guardrails

Generative AI raises many unique security and compliance issues. For example: Who owns the content produced by AI models? What happens to the data used to develop models? What access and controls are needed to ensure the safe, secure and ethical use of this technology?

Unfortunately, we don’t have clear answers to those questions yet. However, the answer is not to hold off on using generative AI or restrict its use.

Since existing security and compliance frameworks do not address these issues, companies will need to develop policies and procedures, often in conjunction with a technology partner, to establish program guardrails. They can also lower risk by initially focusing on internal applications that can be closely monitored and controlled. Finally, they can partner with a reputable hyperscaler with a vested interest in maintaining the highest levels of security and privacy.


  • Develop custom policies and procedures.
  • Choose hyperscaler partners with a vested interest in data security.


  • Rely on existing security and compliance policies and frameworks.
  • Delay the use of this new technology out of fear of change or risk.  

7.    Create new KPIs to measure impact

The goal of a new technology introduction like generative AI is not necessarily to deliver an ROI in the initial phase. Instead, the goal is to prove the technology works in the environment and has the potential to deliver value and impact.

To that end, new KPIs are needed that are specific to the identified use case. It’s important to go beyond traditional metrics to determine the business impact of the program. For example, for an internal knowledge management use case, traditional KPIs might be how many screens or clicks a user needs to go through to find an answer. But with generative AI, the metrics need to be more open-ended and include indicators like the total amount of time spent reviewing information or the number of repositories the person used.

In addition to learning from successful use cases, don’t underestimate the value of failures. Failed programs often offer important insights into technology limits and boundaries. By capturing key learnings from both successful and unsuccessful projects, organizations can get a more complete sense of how to adapt in the future.


  • Set KPIs that focus on value generated over time.
  • Move on when the use case does not work.


  • Ignore the learning potential of failed use cases.

8.    Communicate early and often

When moving from POC and pilot to cross-enterprise implementation, feedback loops are essential. Feedback—especially from the use case evaluation framework—should be built in from the outset to ensure the company can evolve and adapt the solution and the user experience to achieve the highest value. Feedback mechanisms could involve surveys, scoring, instant user evaluations or other feedback tactics.

While communication is often overlooked during major tech program rollouts and change management plans, it’s more important than ever with generative AI, given the amount of attention on this technology—and the accompanying hype and even misinformation.

Specifically, it’s essential to address employee concerns about job displacement, privacy and bias—even if they are not using the technology day-to-day. By communicating clearly and honestly about these issues and, at the same time, emphasizing the practical advantages of the technology for both workers and companies, organizations can help ensure quick adoption and strong engagement with the technology as it’s rolled out.


  • Ensure feedback loops are built-in features of every generative AI system.
  • Communicate clearly and continuously to all employees about generative AI’s impact.


  • Rely on traditional feedback methods to gauge generative AI effectiveness.

9.    Be aggressive about learning and development

The success of the generative AI program will depend, in part, on how comfortable and proficient employees are with the technology. A comprehensive training and development program is needed for people to learn how to use the tools and understand the value in doing so.

The training program should include hands-on, technical courses for those involved in implementation, maintenance and design, as well as general learning opportunities for all employees about how the technology works and how it is being used. It’s essential to engage everyone in this learning and development journey to successfully roll out more use cases and extract more value from the technology.


  • Offer hands-on and general learning opportunities.
  • Integrate the learning and development program with the change management plan.


  • Limit learning and development only to those who will directly use generative AI tools.

10.    Learn through iteration

At this point, every AI program should revolve around learning. That means identifying both what works and what does not.

While most organizations are eager to unlock a blockbuster use case or application, doing so will take time, effort and patience as people get comfortable with the technology and understand where the boundaries lie.

While there is certainly value to be gained from the technology today, the real benefits will be the result of careful, consistent, deliberate iteration based on the knowledge and learning from past programs.


  • Focus on learning and building proficiency.
  • Embrace continuous iteration to drive value over time.


  • Start with blockbuster use cases.
  • Give up on this new, evolving technology.

From sidelines to success with gen AI

Despite being in its early stages, it’s clear generative AI will play an integral role in the future of every industry. Companies taking a “wait and see” approach are putting themselves in a position to fall behind.

While uncertainty is inherent to any new technology introduction, the value of generative AI significantly outweighs the perceived risk. Companies can adopt specific measures throughout the introduction process to mitigate those risks.

Now is the time to get in the game with generative AI—because delaying or denying the use of this tool is likely to create the biggest risk of all.

To learn more, visit the Generative AI section of our website or contact us.

Scott TumSuden

Vice President & Global Managing Partner

Author Image of Scott TumSuden

Scott oversees strategy and growth for Cognizant's Retail division, leading relations with a top Fortune 30 client. A former Fortune 10 tech executive, he drives successful digital transformations in the industry.

Latest posts

Progress with generative AI

Read our blog on Jumpstarting the gen AI journey and visit us at NRF to discuss more.

A lady printing receipt at a retail store

Related posts

Subscribe for more and stay relevant

The Modern Business newsletter delivers monthly insights to help your business adapt, evolve, and respond—as if on intuition