Artificial intelligence technology's advances are at a pace that challenges even the most agile organizations. Deciding what to adopt isn't easy, especially when vendor marketing, messaging, and promises often sound alike. Yet waiting too long risks your organization falling behind and missing opportunities, putting you at a competitive disadvantage and impacting the bottom line. This column shares practical guidelines to help organizations move confidently into the world of customer experience AI and automation.
AI Guidelines for Enterprises
#1: Assess your CX organization's performance to identify opportunities for improvement. Take a broad customer journey perspective that goes beyond traditional organizational boundaries, including the front and back office and all functions, activities, or teams that have previously been overlooked. This holistic approach can easily identify savings of 20 percent to 40 percent while eliminating unnecessary work and enhancing the customer and employee experience.
#2: Identify five solutions/applications that address your organization's challenges and schedule product demos. The AI market is crowded and noisy, and while many CX vendors claim to address the same or similar challenges, their solutions differ widely in maturity, capabilities, and implementation best practices. It's a confusing market, but also a buyers' market. It's good practice to schedule calls with each vendor and provide as much information as you can so they can set up targeted demonstrations. (Establish a non-disclosure agreement (NDA) if required by your company to facilitate information sharing.) When attending the demo, listen for signals such as, "We can customize the product for you," or "We don't have reference customers in your vertical," or "We'll build this together." These often indicate the vendor lacks experience in your vertical or with your specific challenges. However, keep in mind that there are situations where this could be to your advantage, particularly if you need a highly customized solution in a less common vertical.
#3: Draft an AI and automation strategy to guide CX transformation and, where possible, align it with your corporate AI/automation strategy. In general, it's a best practice to start with a strategy and then identify/select the tools and practices that support it. However, since AI is new in many companies and CX organizations, it's often useful to explore current market offerings and planned enhancements and innovation before drafting your company's two- to three-year AI/automation roadmap. Vendors can be a good (albeit somewhat biased) source of market information, as can be industry analysts, thought leadership content, articles, blogs, video, etc. Once the AI/automation strategy is documented and approved by senior management and the corporate AI center or excellence or chief technology officer, reassess and reconfirm the tools and technologies initially selected to ensure they remain the right fit. Of equal importance, carefully review each vendors' product roadmaps and delivery timeframes to make sure they are investing in the areas most critical to your organization. (And keep in mind that vendors are more receptive to feedback during the sales cycle than after the deal is signed.)
#4: Draft a requirements document (request for proposal, RFP) that identifies the functionality and capabilities you need. It's common for companies to draft RFPs with hundreds of questions for vendors to answer. However, as few (if any) vendors allow their RFP answers to be included in the sales contract, this is often a time-consuming effort that slows down the acquisition process. An RFP can be helpful for understanding system capabilities and comparing pricing strategies and costs, but it isn't a guarantee that the solution will work as described. And when it comes to prices, since most AI products are billed based on consumption/usage, actual cost often varies from the price estimates in an RFP. Therefore, DMG recommends that companies start by identifying and documenting their AI/automation requirements. Use this document to get buy-in and agreement from all departments and functions that will use and benefit from the solution. Then, consolidate the requirements into five or six high-level categories that each have three or four questions/items; in other words, draft a three- to four-page RFP that will be the primary driver of a detailed product evaluation and its roadmap. (Since this is what generally happens anyway, it will save your organization a great deal of time and effort.)
#5: Determine whether you want to purchase a platform/framework or stand-alone suite of capabilities. This guideline can be addressed earlier in a selection process but is often easier once you've learned more about the products and their integration capabilities through the RFP process. This is a major area of debate in the CX and enterprise software markets, and the chosen approach should be driven by your company's AI/automation strategy. Independent of the direction taken, it's critical to acquire solutions—AI (ML, genAI, agentic AI), AI orchestration, large language models, digital platforms, CRM, contact center-as-a-service/premise-based automatic call distributors, self-service, knowledge management, analytics, etc.—that are open and can be easily integrated to future-proof your technology environment. The one constant is the need for flexibility so organizations can adapt and adopt new solutions or LLMs as the market evolves.
#6: Make sure you have the appropriate data volume to support each AI project. No matter what the solution, data is the engine driving AI outcomes. It's essential to have access to large, well-sourced data for each AI use case. A general market LLM, for example, might not be the best fit for a vertical specific, CX-focused AI project. AI orchestration and data layers should provide secure, company-wide access to data, supported by a protected knowledge base. This is particularly applicable where personalization and privacy are required, such as healthcare, financial services, and insurance.
#7: Do not base your final decision on the product cost. There are lots of similar products available, and price is a single decision criterion, albeit an influential factor. Almost all AI products are cloud-based and use consumption or time-based pricing models. If a product is successful, as is the case for customer self-service or agent-facing augmentation applications, its volume and cost will increase rapidly as it displaces human agent-related expenses. Purchasing a product that costs less per unit but requires more time and effort to implement and delivers fewer benefits might be much more expensive in the long run. Deep and broad functionality, ease of implementation and maintenance, AI-based with easy-to-use analytics, ongoing innovation, and featured upgrades are at least as important as the product price. For this reason, it's essential to conduct a few financial analyses when selecting a product, including payback period, total cost of ownership, and business metrics, such as efficiency gains, CX/customer satisfaction improvements, and risk mitigation.
#8: Start with one or two use cases and be positioned to expand quickly and cost-effectively. This guideline addresses a couple of important decisions: whether a company should conduct a pilot and how many use cases it should address at one time. Self-service pilots often fail because they target low-volume use cases that cannot deliver meaningful containment rates. Instead of piloting an unimportant or low-volume use case, DMG recommends that companies select one or two high-volume use cases that will test most aspects of the solution and vendor: everything from its low-code/no-code development environment, openness, LLMs (and the ability to switch between them, as needed), product functionality, agentic-AI capabilities, AI-enabled reporting, implementation best practices and speed, transparency, and vendor reliability, expertise, support, and responsiveness. However, even in situations where conducting a pilot is not the right approach to getting started, prospects must retain the right to cancel their contracts without penalty if objectives are not met. And if the system performance cannot be replicated in subsequent use cases as promised by the vendor, companies must also have the right to cancel their contracts without hassle or fees.
These guidelines make it clear that investing in and implementing AI-enabled solutions comes with considerable risk and effort, but the cost of doing nothing and waiting for the market to settle is far greater. Organizations that approach AI initiatives the right way are highly likely to realize substantial and measurable benefits: scaling and personalizing the customer experience, boosting satisfaction, and enhancing the employee experience and operational efficiency. It's not easy, but with the right partner, solution, and guidelines, it&'s well worth the effort.
Donna Fluss, founder and president of DMG Consulting, provides a unique and unparalleled understanding of the people, processes, and technology that drive the strategic direction of the dynamic and rapidly transforming contact center and back-office markets. As the foremost analyst and visionary dedicated to the contact center and back-office markets, she has provided expert guidance for more than 30 years to technology leaders as well as disruptive newcomers, investors, and companies that want to build next-generation AI-enabled contact centers. She can be reached at Donna.Fluss@dmgconsult.com.