- AI Integrated Newsletter
- Posts
- What is An AI Strategy and Why Every Business Needs One
What is An AI Strategy and Why Every Business Needs One
This article discusses what an AI strategy means, the different types of AI strategies that you should know about, and how as a leader you can get started with an AI strategy.
What is an AI strategy?
An AI strategy may seem like a complicated business-speak, but it’s simply a vision or high-level plan for integrating AI into the organization, such that it aligns with your broader business and automation goals. This high-level plan can be a:
Product-level AI strategy (I often call this AI product strategy)
The granularity of the plan is often inversely proportional to the magnitude of the vision. So, the bigger the vision, the broader the objectives until each objective is broken down into a roadmap for implementation.
Why Do Businesses Need an AI Strategy?
So now the question becomes, why does an organization need an AI strategy? Why can’t you just dive into implementation or make an off-the-shelf purchase with little planning?
As Dale Carnegie says, “An hour of planning can save you 10 hours of doing” and with AI it can save you much more than that. It can save you months in setbacks and unnecessary spending. Let’s look at how an AI strategy can be helpful.
Product Level AI Strategy
At the product level, an AI strategy is needed so that you have the foundational building blocks to ensure that an AI project can progress from an idea to a useful tool to creating meaningful results for the business.
The building blocks at this level, could be the data that’s needed to support the initiative, the personnel who will build your data pipelines and develop the relevant models, the metrics to track success, or the competing off-the-shelf solutions that you should consider for integration. Without this level of detailed planning, you’re bound to run into roadblocks that will keep setting you back months at a time.
Elements of an AI Strategy at the Product Level
For example, in early 2020, one of our healthcare customers wanted to develop an AI tool to automatically extract specific information from clinical records, primarily containing free-form text. They had been doing this information extraction work manually for years to obtain reimbursement from insurance companies.
But as we were digging deeper into the project, shockingly, we discovered that none of that data was available. What happened?
As part of the company’s manual process, they copied data directly from the clinical records and pasted them into third-party forms for reimbursement. Unfortunately, once the forms were submitted, the company kept no records of the information that they copied over. This meant—no historical data for AI. As a result, the company had to first augment its manual processes with data collection in mind. This set them back six whole months.
This is a common problem for many companies, where the data to build the AI product is often nonexistent, is in the wrong format, or is yet to be collected. This is not just data for model training but also data for evaluation and learning more about the problem that you’re solving.
The building blocks for every AI product can look slightly different depending on whether you’re customizing an existing solution, building from scratch, or buying an off-the-shelf solution.
Business Unit Level AI Strategy
At the business unit level, an AI strategy is needed to ensure that you’re pursuing AI opportunities that will give you the biggest improvement in your bottom line, productivity, or decision-making.
Some of the drivers of those decisions include the severity of the problems that you’re addressing, data availability, infrastructure readiness, and availability of shared resources, and in many cases, you’d have to ensure that the selected initiatives align with the business’s broader objectives.
To prevent you from tackling the wrong initiatives first (i.e., low-impact problems, no data problems, no buy-in problems), requires planning, which can help you be very strategic in implementing AI in your business unit.
For example, if a new problem has no available data for development and evaluation, you may want to start the process manually first. And by doing this, you’re helping your future AI initiative in the following ways:
Establish a baseline performance, which you can later use for A/B testing
Generate high-quality data
Get a better handle on the problem
Leverage domain expertise in developing your AI solution in the future
To sum up, a business unit level AI strategy is primarily to create a logical AI implementation roadmap to maximize impact.
Organizational Level AI Level Strategy
An organizational-level AI strategy is all about getting your organization as a whole ready and prepared for AI. This ensures that you’re not just running one-off pilots in an isolated business unit, but teams across the organization have AI on their radar, know how to tap into the necessary resources to get AI projects going, know how to spot AI opportunities, and have a base level of AI understanding.
Getting an organizational-level AI strategy up is more than just finding AI opportunities to go after in your organization. It’s also about identifying all the readiness gaps from the perspective of:
Budget (B)
Culture (C)
Infrastructure (I)
Data (D)
Skills (S)
All of this is so that a company can take an AI idea from conception to implementation to reaping measurable benefits—all with minimal friction, repeatedly.
In The Business Case For AI, I talk about the five preparation pillars for AI, referred to as B-CIDS (pronounced, “be kids”), which encapsulates the elements we discussed above.
The 5 AI Preparation Pillars From The Business Case For AI for an Organizational-Level AI Strategy Development. Data readiness is the biggest preparation pillar, followed by cultural readiness.
Each pillar is critical, with its own set of requirements for building an AI-ready company.
For example, under cultural readiness, there are six elements to consider to prepare a company culturally for AI. This is due to fears around the technology, the ethical dilemmas, the uncertainties that AI brings, and the need to be highly data-driven.
If you’re not culturally ready for AI, you may experience resistance internally to the adoption of the technology, your employees may not know how to collaborate effectively on AI projects, and the risks that come with AI, including mitigation strategies, may not be well understood.
Another critical pillar, data readiness, is all about ensuring your data warehousing is strong and that you’re collecting data from the daily running of your business and are digitally ready to leverage your data stores. Essentially, this is all about getting your data strategy up and running.
Building up each readiness pillar takes time and a whole lot of planning. And a plan alone is not enough, you need an ACTIONABLE plan—something you can implement.
To facilitate the planning process, I’ve outlined The Jumpstart AI Approach in The Business Case For AI for getting started with your AI strategy using actionable, short-term strides. In a future article, we’ll explore the approach.
AI Startup Strategy
An AI startup is either selling an AI tool or technique that you can use within your applications (e.g., Perspective API), or AI is a big part of the startup’s products and services (e.g., Jasper.ai, the content writing assistant).
If you’re looking to build a startup with AI as the primary driver of your business, then an AI strategy for your product is a must.
Depending on whether you’re developing the AI solution from scratch or using an off-the-shelf model to power your applications, the priorities can look slightly different. However, having worked with multiple AI startups, I can say that these are some common concerns:
Model/Technology
Are you leveraging an open-source solution, buying, or building from scratch?
Who will perform the model development or customization, testing, and integration?
How will the solution integrate into your software or how will users access the solution?
Performance
How will you assess the quality of your tool in solving a broad spectrum of unseen problems?
How do you ensure the quality of your tools stays consistent with time?
Monetization
How will you monetize your tool?
Safety
How do you ensure that your tool produces the expected output?
How will your tool fail gracefully instead of annoying users?
Messaging
What exactly are you selling? The technology and quality behind the technology or the product’s benefits to the end user?
Data
How will you start collecting data from day one to improve models, develop new models, etc.?
Adoption
How will you ensure that you’re addressing fears around AI and helping new users become open to using your solution?
Feedback Loop
How would you collect feedback from users to improve your offerings continuously?
As you’ll notice, this has elements of an AI product strategy and some organizational-level AI strategy, with some monetization and marketing strategy. AI startups must prioritize their messaging, monetization strategy, quality consistency, and feedback loop, as AI is at the front and center of their business.
This is not to say that these elements are not critical to non startup AI products that augment existing workflows. It’s just that the priorities in those scenarios could look very different.
For example, for a large company seeking to enhance a workflow, a 50% boost in productivity through AI may suffice regardless of the model’s accuracy as long as it meets the task requirements. But for an AI startup selling an AI tool, poor or even just “acceptable” accuracy can destroy a buyer’s trust.
AI startup strategy has elements of an AI product strategy and organizational-level AI strategy.
Getting Started With Your AI Strategy
The first step in getting started with your AI strategy is to understand what it is, why it’s important, and if you’re considering an organizational, business unit, or product-level strategy. The scope of your AI strategy will determine what preparation elements to focus on.
At the product level, we saw that we’re primarily focused on the building blocks related to the specific AI project, such as data and personnel.
But at the organizational level, we’re focused on getting the entire organization to have AI on their radar and be able to repeatedly take an AI idea into production and deliver value for the business with minimal friction. This takes much deeper and lengthy preparation along different preparation pillars.
For an organizational or business unit level AI strategy, you can use The Business Case For AI, for concrete steps and frameworks to develop your AI strategy. For a product-level AI strategy, you’ll have to think about all the elements outlined in the first image.
Reply