A Framework for Enterprise AI Success: 8 Framework Elements to Guide Generative AI Projects

A Framework for Enterprise AI Success

Generative AI is changing the way organizations build software and interact with technology. While Generative AI has the potential to revolutionize businesses, it can only deliver returns if implemented strategically and tactically.

Tactical investments will keep an organization competitively viable. Many of the current use cases focus on basic cross-industry tasks. Implementing them is necessary to remain competitive, but returns will prove hard to capture given the distribution and embeddedness of the work, and the more strategic investments must focus on areas of differentiation. Generative AI offers organizations the ability to create transformative tools for their operations and their customers, but it also provides an opportunity for disruptions by start-ups that see potential or are more willing to take the risk in an area than an incumbent.

Regardless of the motivation or the level of investment, a few principles need to be considered to establish a foundation for successful generative AI implementation. The following outlines some of those key ideas. One blog isn’t going to guarantee a successful GenAI project, but this framework will help ensure the project has a fighting chance.

1. Define Clear Objectives and Align with Business Goals

• Clearly define what you want to achieve with AI.
• Identify specific business challenges that AI can address, understand stakeholder pain points, and outline success criteria. Look also for areas of innovation that are usually outside the traditional automation discussion that may have been written off as too hard or that the technology isn’t ready. With generative AI, those use cases will likely be in play.
• Ask:
  • What specific pain points are we trying to address?
  • How will AI optimize current processes?
  • What quantifiable improvements are we expecting?
  • How does AI contribute to our overall growth?
  • How can AI help deliver on our vision?
• Ensure AI initiatives support broader business aims. Productivity isn’t the only business aim, and the commoditization of basic AI-powered tools will make it hard to capture value from customer service teams writing customer responses or searching for information faster. Transformative AI projects will not only cost more; they will take more time to define how they fit into the business model, how they contribute to results, and what the organization will need to do to leverage the opportunity.
• Your AI strategy should be aligned with your overall business strategy. One of the issues organizations face is that they have a near-term business strategy that will be difficult to align with longer-term AI-driven possibilities. Too often organizations take the admonition to align with strategy using their current strategy. AI likely requires a strategy review and perhaps a rewrite. If AI is seen as incremental toward achieving current strategic objectives, then that is how it will be perceived and implemented, and incremental improvement results will follow.

2. Assess Feasibility and Resources

• Evaluate technical feasibility. Most organizations don’t know what generative AI can do, because even its developers aren’t sure. Many proofs of concept (POC) are finding it hard to move to the next level, not just because the questions generative AI is being asked to solve in a POC don’t test the edges of the domain knowledge, but because scaling up isn’t just a matter of quality of the code—it requires more sophisticated development than many organizations are used to. Rolling out an app with menus and taps requires robust reliability and accuracy, but a generative AI system also requires a deeper understanding of user intent and a model of behavior that may not be bounded by the application, which makes testing much harder, and feasibility more difficult to determine in the POC because many of the interactions won’t be known early, and perhaps even after roll-out.
• Evaluate financial feasibility. Examine if the organization can afford the investment and which model will best serve the development of a solution and its long-term operation and maintenance. Keep in mind that AI infrastructure designs continue to evolve, so part of the investment must involve monitoring developments and evolving the plan, as well as its financial implications as the technology shifts.
• Assess technological infrastructure, availability of skilled personnel, and project budgets.
• Plan for training and support to bridge the AI fluency gap between leaders and workers. CIO recently reported that almost three-quarters of employees blame their employer for not giving them time or access to AI training. That needs to be fixed. AI is developing rapidly. It has no fixed curriculum that can certify readiness one week with an expectation that the training will remain relevant a week later. New models, new approaches, and new vendors all mean a rapidly developing field that requires regular updates in order for knowledge to remain relevant.

3. Select the Right AI Solutions and Vendors

• Choose secure and reputable AI providers. Choosing a secure and reputable AI provider is crucial for protecting your company’s sensitive data. Almost two-thirds of knowledge workers and business leaders cite this as a top concern when it comes to generative AI. Look for a provider with a strong security approach that includes enterprise-grade attestations and regulatory compliance. Additionally, prioritize vendors who focus on data ownership and policies that focus on safeguarding data.
• Ask vendors:
  • How will the system benefit our specific needs?
  • How difficult is it for employees to learn?
  • What training and support is provided?
  • Is there a demo or trial period?
  • How do you address ethical considerations?
• Select solutions that can be easily embedded into existing workflows. Any AI solution that won’t work with existing enterprise software will create delays, and if they are too long, the underlying idea of how the AI will work in the solution may need to be reconsidered if the technology has already moved on.

4. Build a Data-Driven Foundation

I can’t emphasize enough the importance that data and AI must be considered co-equals in generative AI projects—perhaps even with data being the more important of the two because enterprise AI systems without enterprise data are just internal implementations of someone else’s data (perhaps some of yours, but you will probably never know how much or how accurate).

• Align AI and data strategies with business objectives.
• View data as the fuel for AI.
• Ensure high-quality, unbiased data is used to train AI models.
• Establish data quality control processes with rigorous collection, cleaning, and validation.

5. Foster Policies and Practices that Encourage Innovation and Collaboration

This group of tasks has nothing to do with technology and everything to do with its ultimate success or, perhaps, more importantly, its impact. AI, like many technologies, however, brings with it a promise and a threat. It promises to offload mundane work, but it also threatens the continuity of work and experiences for those involved in its creation.

• Cultivate organizational practices and policies that embrace innovation and continuous learning.
• Encourage cross-functional collaboration between data scientists, AI leads, business leaders, and enterprise IT teams.
• Promote knowledge sharing and provide ongoing training and development opportunities.
• Develop data literacy across the organization with tailored training programs.
• Create lab spaces that are safe havens for experimentation.
• Realign incentives and permissions in work relationships to allow people to more fully participate and get rewarded for their contributions.

6. Ensure Responsible and Ethical AI Implementation

While we often think of ethics as universal, for most AI systems, ethics are contextual. Each organization needs to define its unique version of ethics for its industry and stakeholders. General platitudes and references to ethical frameworks will not suffice in the long run. So, while this section is general, think of each area as requiring an internal definition that aligns with business objectives, which, in some cases, may cause the organization to be self-reflective on its strategy, not just its use of AI.

Regardless of the nature of the internal conversation, the most important aspects of ethics involve sharing the framework and executing within it. Violations of ethical behavior in the use of AI are no less forgivable than unethical behavior in any other aspect of life. Do not wait for legislation or regulation, and do not put all the accountability for AI ethics on vendors. Each use case involves ethical choices by those deploying the solution, not just those creating the technology or regulating the industry.

• Promote confidence and address ethical and security risks.
• Establish clear guidelines for ethical use and create trustworthy practices.
• Consider potential biases in data sets and models.
• Implement mechanisms for human oversight, auditing, and addressing unintended consequences.
• Prioritize transparency and be open about how and when AI is being used.

7. Measure Success and Demonstrate ROI

First, organizations need to plan on how to measure the success of AI projects. Many generative AI projects may not pay off in quick productivity wins, and those that do, may deliver value at levels below typical enterprise KPIs. The best organizations will include finance in the design of measurements that can track the value of generative AI solutions over long periods of time. Ideas like The Serendipity Economy offer insight on how to track long-tail value realization and how to recognize atypical value models that arise from the use of technology.

• Set clear, quantifiable metrics for success early on.
• Measure AI’s impact on key performance indicators, including cost savings (with efficiency gains rolled in) and revenue growth (with customer satisfaction rolled in). Avoid fuzzy metrics that can’t easily align with cost savings or revenue growth.
• Don’t solely rely on traditional ROI metrics; consider factors like agility, competitive advantage, and risk tolerance, all of which, at some point, will align with cost savings or revenue growth. Use tools like the “value tree” to define the scope and impact of AI.

8. Embrace Continuous Improvement

Ideas behind “continuous improvement” have been alluded to in many other items on this framework. For generative AI, continuous improvement is more important than in many other technology domains because of its rapid evolution. There is no point at which an AI application becomes a stable production system that requires only minor technical improvement.

It is also increasingly unlikely that traditional IT Maturity models will be able to incorporate AI because it will remain, at least for a number of years, in a constant state of becoming. Even systems that might be considered hands-off require monitoring to ensure that the generative AI components, some of which developers may not have control over, continue to perform as needed (note, I did not say, “as expected” because responses change with each prompt but remain functionally equivalent, which is new for IT developers).

• View the AI journey as iterative and be prepared to adapt and evolve with the technology.
• Continuously test, learn, and refine AI models based on data and feedback.
• Stay informed about emerging trends and advancements in the AI landscape.

By following this framework, enterprise AI buyers can increase their chances of successful AI implementation and realize the transformative potential of AI.

For more serious insights on AI, click here.

All images created by meta.ai from prompts written by the author.

Did you enjoy a Framework for Enterprise AI Success? If so, please like the post and share it on social media. Click a sharing button for easy sharing. Have a question or comment? Please engage in the comments section!