This article was a request from one of our Catalyst Advisory Program members. The Catalyst Advisory Program is an application-only business growth coaching program for AI consultants and AI service providers. The program helps AI consulting and services leaders win more deals and deliver more client value. Members receive one-to-one advisory, group coaching, and proprietary Catalyst AI best-practice frameworks. Learn more or apply at: emerj.com/catalyst.
Last week in our Emerj Plus best practice guide article, we covered the keys to selling AI services into the enterprise – and this week we’ll focus on AI procurement and adoption from the perspective of the enterprise leader themselves.
At Emerj, much of our work with enterprise innovation and strategy leaders – professionals who are in charge of a broad swath of technology adoption and deployment projects across different parts of the business. Because innovation and strategy titles often have a mandate to be forward-thinking, these VPs, directors, or “heads of” innovation and strategy are often among the most informed members of the leadership team when it comes to understanding AI.
In this article we’ll cover five steps to successful AI deployment for enterprise innovation leaders:
Note: I’ve created these steps for innovation and strategy leaders within enterprise companies. Chief information security officers or heads of engineering would have different areas of emphasis, and different steps to follow.
This article already presumes that the AI buyer in question here is someone who understands what AI can do, how it is adopted, and fundamental AI use-cases (i.e. Executive AI Fluency, the first element of Emerj’s “Critical Capabilities” AI maturity model). For enterprise leaders in charge of AI strategy and procurement should master a conceptual understanding of AI before following the ordered steps in this article.
We’ll explore the phases in order, beginning with the ability to assess organizational AI maturity.
1 – Assess Organizational AI Maturity
If we’re considering AI deployments seriously, we’re thinking about where we want our company to be, or how we’re going to transform. We can’t think seriously about transformation without a thorough understanding of where we’re starting from.
Because “AI maturity” or “AI readiness” are broad topics, defined differently by different experts.
At Emerj, we’ve simplified the elements of AI maturity for functional business leaders in our Critical Capabilities model (seen at the right). This model would be different for machine learning engineers or IT leadership, but for CEOs, innovation leaders, and most non-technical VPs and Directors, this model represents the critical areas of improvement required to successfully deploy AI.
Skills – particularly executive AI fluency – are a critical first step in assessing AI priorities and allocating resources to AI in an effective way. Skills can be broken down into:
- (1.1) functional leaders that understand AI
- (1.2) data science skills and talent
- (1.3) cross-functional AI teams
An enterprise would obviously need to have a pool of workers with the skills and knowledge to work with AI at a high level such as data scientists and data engineers. They would be in charge of cleaning and organizing data to train AI and come up with a strategy to deploy AI to the benefit of the enterprise.
However, the technical team would need to work within the parameters of the goals of an enterprise. They would be a cross-functional team in which they work with non-technical teams (marketing, sales, HR, operations, etc.) to deploy AI into the business model with an eye to ROI. The lack of this concerted effort is a big reason why AI is hard to adopt successfully in the enterprise.
On top of the cross-functional teams should be functional leaders with a conceptual understanding of AI capabilities, basic terms, and representative use cases. Functional leaders could be a CMO or a Head of eCommerce and are generally non-technical executives in charge of the overall AI strategy. Without leadership who can (a) identify viable and useful AI applications, and (b) set realistic expectations for AI ROI and deployment considerations, data scientists and cross-functional teams and left to work on doomed projects.
Enterprise and business culture is an academic study unto itself – in the context of AI maturity, we emphasize a limited number of required cultural changes:
- (2.1) willingness to experiment and iterate
- (2.2.) value data, and
- (2.3) value cross-functional collaboration
AI use-cases are nascent, AI systems are iterative and probablistic (unlike IT systems that can do a job 100% of the time, an AI system will often perform a job within a specific performance range, and will require ongoing monitoring). This means it should be willing to constantly experiment with AI by collecting and assessing data and updating algorithm parameters accordingly without any assurance of when there will be light at the end of the tunnel in terms of ROI.
That might seem a lot of uncertainty, but that is why functional leaders need to have a conceptual understanding of AI. If they have a better idea of the time and resources involved in an AI project, they’ll be more willing to experiment and iterate to achieve ROI.
Connected to experimentation with AI is the availability of data. Data is the key to developing AI for successful deployment, so it follows that without proper data, there can be no AI deployment.
To achieve AI readiness, an enterprise should give due consideration to maintaining, storing, and accessing the data in a usable format. There should be protocols for categorizing, maintaining, and using existing data assets as well as acquiring new ones. Valuing data involves deliberate efforts, including how it’s going to be used and how that informs the manner of storage and access.
It should be noted, however, that experimentation and data management cannot happen without effective collaboration across the board. Each team would have their individual goals, but it should drive them towards a common one. Effective collaboration would also identify areas in which there are gaps such as the need for additional people and other resources. AI deployment success hangs primarily on the ability (or inability) of different teams in an organization to work together to achieve a common goal.
Aside from human resources, an enterprise would need two core resources to deploy AI:
- (3.1) data access
- (3.2) in-house AI playbooks and guides
They need to know the data they have and the data to which they need access for AI deployment. A finance company, for example, would need data on the performance of portfolios and financial markets, fraud detection, and regulatory compliance, and a way to access them.
When data access is not a problem, the enterprise should assess the data quality for leveraging it to produce the desired results. This often means experimenting with the data in small projects to figure out what facets of it are valuable to them as well as identifying the correct format and architecture to make the data accessible and useable for a given project. This last part is important, as the wrong data structure renders the data useless.
Another type of resource you need for AI deployment is in-house AI playbooks and guides, which are often overlooked. An enterprise can deploy AI successfully if there is a framework or guide of best practices (so far) for maintaining existing AI projects. The guide or playbook would include information on the teams needed, data access, and the types of integrations that need to take place. Playbooks might include:
- Protocols for pulling subject-matter experts into cross-functional AI teams.
- Checklists for data access, harmonization, or security (a variety of these lists might be created for different individual data types or silos).
- Measurement metrics to be used for different kinds of machine learning projects, paired with which kinds of workflows they’re best suited for measuring.
- Rules of thumb for estimating the number of data science team members who may need to be dedicated to a specific initiative or project.
Enterprise innovation leaders would need to create these guides or playbooks by retaining what they learn as they go through the phases of AI deployment. However, it is important to note that these guides will have to evolve as it might take years to find the best way to run AI projects. These playbooks often fall under the purview of an AI “center of excellence,” but can be handled by innovation or digital transformation leaders.
Retaining learnings is something that successful AI adopters do well, and something that most enterprises neglect as they strive for near-term results without focusing on building competence and maturity.
2 – Align AI Potential to Near and Long-Term Goals
Experienced AI vendors typically anchor the potential of their product as a solution to an immediate problem without going in too deeply into the complexity of AI to avoid scaring them away at initial contact. However, as we have already discussed, enterprise innovation leaders should have an adequate level of understanding of AI to align its adoption with near and long-term business goals as well as AI readiness, so this avoidance of real-world applications of AI should no longer be necessary.
Adequate executive fluency is critical in choosing AI based on the level of maturity in different departments that will cover not only 3 to 5 year goals and current thrusts and initiatives, but lead to strategic differentiators for competitive advantage. It will point to use case opportunities and areas of the business where AI could fit into different workflows.AI could be a game-changer in any industry if natively aligned with business goals.
In effect, the role of the enterprise innovation leader is to develop a digital or AI transformation vision. It is forming and molding a vision of the direction to where the company is going to generate ROI. It could be through developing frameworks for a checklist of typical “North Star“ business objectives tied to each industry sector and a starter list of AI projects and industry that have shown higher statistical odds of successfully from pilot to production.
In general, however, aligning AI is not an easy task, as historically, most AI projects don’t achieve an ROI of any kind – and it has little to do with data or data science talent. In many cases, this is because measurements of success – and indeed the Ai transformation vision — were established incorrectly.
It falls on the enterprise innovation leader to prevent that from happening and avoid setting an AI project up for failure. Realistically determining where the enterprise is going with AI is crucial for success in its deployment.
3 – Matching AI to Company Needs and Maturity
As an enterprise innovation leader – if you know where your company stands in terms of AI maturity, and you have a firm understanding of the near and long-term technology priorities of your organization, you can begin developing potential AI projects that are (a) accessible given your resource and talent constraints, and also (b) align best with the objectives and priorities of your firm.
Identifying the Full Range of AI Opportunities
The full range of available AI use-cases can be drawn from a variety of sources. Most enterprise leaders use the press releases of their largest competitors to be a representative set of viable AI projects – but nothing could be further from the truth.
Companies exaggerate and promote AI initiatives that will make them look good to their customers and investors, and they conceal and downplay AI initiatives that don’t make them look good to their customers and investors.
Emerj’s AI Opportunity Landscape research maps the full capability-space of AI in across sectors – but firms without market research budgets may collect AI use-cases from a variety of sources including:
- Listed AI use-cases on the websites of AI vendors in your industry
- Conversations with peers in similar roles and industries (specifically, peers with experience procurement and deploying AI solutions)
- Look for the use-cases of AI vendors who have taken substantial investments from the largest enterprises in your industry (retail bankers might look at the investment activity of Wells Fargo or Citibank, for example)
- Among other sources…
If you simply avoid using competitor press releases as your baseline, and draw from other sources in order to flesh out a use-case landscape that matches your business processes and business objectives, you’ll be ahead of the majority of enterprise leaders who see a much more myopic range of use-cases.
Scoring AI Opportunities by ROI Potential, Ease of Deployment, etc.
A list of AI opportunities must be scored and sorted in order to determine what might be best. At Emerj, we use our own set of proprietary scores (including “ROI Potential,” “Ease of Deployment,” and more), but enterprise leaders can choose to use their own scoring criteria, which might include:
- Alignment with the firm’s existing digital transformation vision
- “Fit” with existing levels of AI talent, data infrastructure, or other elements of AI maturity
- A rough estimate of cost or time to deployment
Sorting and scoring AI opportunities allows enterprise innovation leaders to bring only the very best AI projects to their leadership teams – along with strong reasons behind the projects selected. We recommend that innovation leaders get a clear sense of the priorities of their buying stakeholder (whoever is responsible for cutting the check), and to factor those elements in directly to the scoring and sorting process for potential AI projects.
Our three ROI types are another recommended strata for comparing and prioritizing AI projects, and we cover that in our next section:
4 – Make the Business Case for AI
Once an enterprise innovation leader knows which AI projects suit the company’s goals, budgets, and AI maturity best, they have to convey the value of these proposed AI initiatives to leadership.
The vast majority of the time, these buying stakeholders (often executives) have much less context about how AI works and what AI is capable of, and so communication is critical in order to not only sell the near-term results of AI, but to set expectations about what it will take to genuinely turn the initiative into value for the company.
Conveying the ROI of AI
Countless millions of dollars have been wasted on AI projects because poor project selection – but the same amount might be wasted because of an over-emphasis on immediate, near-term AI ROI. It is critical for innovation leaders (or an AI champion or catalyst within an enterprise) to emphasize strategic and capability ROI as well.
- Measurable ROI – Financial ROI such as improved efficiency, improved revenues – or proxy metrics, such as reduced time in a particular workflow, improved customer service scores (self-reported by the customer via survey), etc.
- Strategic ROI – Factors that bring the company closer to their strategic aims – such as their 3-5 year goals, their key thrusts or initiatives, their existing digital transformation vision, etc. How will your AI project tangibly bring your organization closer to its longer-term strategic goals?
- Capability ROI – Factors that develop Critical Capabilities (see image above), and move a company closer to AI maturity. Measuring AI capability is a topic for its own article and will be covered later, but recording before-and-after progression various AI maturity factors is important for conveying capability ROI to leadership.
If capability ROI (i.e. the development of AI maturity factors, Critical Capabilities) isn’t addressed, then the challenges that arise in AI deployment (issues accessing the data, issues getting cross-functional teams to collaborate, issues encouraging a culture of iteration) will seem to be frustrations and hurdles, rather than as opportunities to improve the company’s skills and culture to enable future transformation.
Capability ROI can be a touchy topic because it implies challenge and change, but if a project will require AI maturity to be developed (and most AI initiatives will, to one degree or another), responsible enterprise innovation leaders should address this requirement head-on, lest their project get cut or canceled due to what buying stakeholders view as “setbacks.”
Presenting the Business Case to Stakeholders
Putting together a business case that will persuade stakeholders to put their money done requires a lot of research and data gathering as well as constantly interfacing with functional business leaders and subject matter experts to make a compelling case. This is particularly important if you are consulting with an enterprise rather than being an in-house resource.
Many companies with whom we work in a consulting capacity indicated that they needed the market research and data we provided to make their case to internal leaders to get them to sign off on the projects. The charts, the quotes, and the primary market research insights provided a strong argument for AI ROI to stakeholders, which is obviously what you want.
We’ve covered the three elements of a successful AI business case in a previous Emerj Plus article.
Building Stakeholder Buy-In
As an enterprise innovation leader, you need to take advantage of your access to the check cutters, functional business leaders, and subject matter experts to interface with them and get their support. Failure to build stakeholder buy-in to your project will make it very hard for you to get the support and resources you need for effective AI deployment.
You need top management on board for financial and resource support, but you also need to get the commitment from functional business leaders and subject matter experts to discover where the opportunities are for AI deployment. The research you make will provide you with strong references to put together an argument for the necessity of an AI initiative, and not just as something nice to have.
If we want to bring a project to bear and ultimately present a project, they need to get them all signed up. You want allies in areas where you plan to apply AI to help carry things forward, and ideally with some level of executive AI fluency.
5 – Guide Practical AI Deployment
As an enterprise innovation leader or AI consultant, you need to be the catalyst of education and set expectations about delivering enterprise value and guide practical AI deployment. This is an important part of AI deployment, because not all AI projects are going to be of benefit to an enterprise.
Setting Expectations About Delivering Value
As we discussed in the previous section, a key factor in making the business case for an AI project is to demonstrate ROI. However, there are many misconceptions about the proper way to predict the return on investment and what can and can’t be done with a particular project.
In practical terms, delivering value with AI projects comes down to its relevance and usefulness to the enterprise, cost, and integration issues. However, you first have to ascertain if you can determine for sure that AI will give you the results you need and/or want.
All of these are difficult to pin down in many cases because there is so much going on with AI. All the hype regarding the capabilities and desirability of AI has made it challenging to get down to brass tacks.
The fact is adoption of AI will not make sense for all enterprises. Too many companies have sunk a lot of money into AI projects and failed to get the results they needed. You and your teams have to get smarter and more capable to know the difference between a good and stupid AI project so you can set realistic expectations and avoid failure.
Building AI Maturity Through AI Projects
Launching an AI project for the first time will have a lot of growing pains even if you follow this guide to the letter. AI is largely an unknown element, and it would be impossible to iron out all the wrinkles prior to deploying it.
However, you can use small AI projects as a way to build AI maturity and work your way up to larger projects. You can think of it as a beta test for weeding out the bugs and finding out ways to effectively deploy AI. Eventually, you will increase your AI fluency and be able to make fairly accurate predictions of AI ROI and adoption in both the near and long term.
Expanding AI Initiatives Through Iteration
Once you have established realistic expectations, built AI maturity, and have executive permission to leverage AI applications for initial pilot projects, the next step is to expand these initiatives through iteration. This would involve going into tertiary projects by expanding future AI capabilities for particular projects, assuming everything goes well the first time around. You can sell this idea by providing a statement of work and a list of activities and deliverables needed to guide the strategy based on changing AI goals and learning that demonstrate how it runs counter to typical AI projects.
The value of expanding through iteration is that you and your team are growing with the capabilities of the AI technology in which you invested. You are pushing the envelope in a smart way because you already know it works to a certain point. You are taking on less risk by simply expanding on its capabilities and maybe tweaking it a little as you continue to learn more about the technology. This would not only help you deploy AI successfully, but develop it further to give your enterprise a competitive advantage.