Today, AI is an integral part of modern CRM platforms and Salesforce is increasingly incorporating generative AI, intelligent agents, predictive analytics, and automation to assist with sales productivity, customer support and many other elements of modern business operations.
However, deploying AI functionality is not just a simple feature enablement; the more significant hurdle for consultants, developers, and technology staff is the creation of a viable, secure and accurate AI model aligned to true business requirements.
There are a number of factors, and these are seven of them:
- The State of Data
Reliable AI starts with reliable data.
Many Salesforce instances hold years of history in the form of cases, customer records, leads, activities etc that may over time become inconsistent, fragmented, outdated or duplicated. Before implementing AI, all of the following should be investigated in relation to:
Data completion rates
Duplicate data entries
Data inconsistencies
Data governance policies
Data integration quality
Data access rights and levels
The impact of unreliable data can lead to useless AI output.
- Integration With Existing Systems
Salesforce is rarely a standalone part of the enterprise: integration with other applications such as ERP, marketing systems, customer support and a customer data warehouse are commonplace. Any AI implementation within Salesforce will therefore need to be considered against a wider integration profile and will require movement of data in and out of the system. While conceptually an exciting AI innovation, technically, an AI application with no real-world business context, nor ability to integrate with upstream or downstream applications, simply is not valuable.
- Data Security & Privacy Issues
Introducing another intelligence system raises the stakes for security: consultants will need a clear understanding of what data the AI model can access and what access to it will be granted for those individuals who engage with it (as opposed to all individuals within an organization). This is an important factor to consider for the sensitive customer and business data routinely held within CRM. Security need to be built-in at the design phase and not bolted on post implementation.
- AI Hallucinations and Reliability Issues
A particular aspect of generative AI can be the propensity for a machine to create convincing sounding, but factually inaccurate output. When using AI for customer interactions or to create text documents this is inherently problematic for the organization, and measures to implement checks and balances and appropriate verification of information produced will be important before the AI tool can be widely trusted. The most positive use of AI in these instances should be as an assistance tool, to enhance the work already being done by individuals, rather than automate decision-making that should be done by humans.
- Employee Adoption and User Buy-In
AI implementation projects should also be seen as a change management project too; employees are still often used to old workflows, new and complex processes can be slow and frustrating, particularly those dependent on new technology. Trust has to be established through user training and adoption is more likely if people perceive new tools as being easy to use, helpful and a benefit, rather than a complicated tool of another sort.
- Business Value Measurement
The mistake frequently made in business transformation is to measure progress on the basis of implementation: for example: how many features has each individual been trained on. A much more effective method is to measure progress on the basis of solving business challenges. Is this AI implementation solving a problems such as:
Higher sales productivity?
Improved lead conversion rates?
Reduced customer support response times?
A reduction in manual and time consuming tasks?
Better customer retention or higher overall customer satisfaction?
Decreased operational costs?
Without such measurable objectives it will not be possible to evaluate the business value of implementing the AI feature.
- AI-Powered Communication
AI can already perform an abundance of tasks in the Salesforce platform, however the future of such implementation also lies firmly in improving communication-with modern organizations generating truly enormous amounts of content that require assistance: customer interactions; interpersonal communication; collaboration; business documents; email responses etc.
CommCon AI is, for example, developing approaches in this space and you can find more details here.
Conclusion
There are a host of data quality, integration, and design challenges associated with successfully implementing and managing an AI capability with an existing and dynamic Salesforce implementation. However, in my estimation the challenge in its entirety may well be the understanding of where to apply AI across the business first and foremost. There needs to be an established and defined role for the AI system and it must have the relevant and contextual data input to perform.
Instead of asking "Where can we add AI?" a stronger stance might well be, "Which is the business problem that we need to solve most urgently, and can an AI add significant value to how we approach it?" That can make all the difference between an installation that adds prestige and one that yields real business benefits.
Top comments (0)