Artificial intelligence (AI) is changing how organizations operate and directly impacts how technology is being used to support and provision services. Although there is much to be said about how managed service providers (MSPs) can use AI and generative AI (GenAI) internally to improve their service delivery process, there is also much to be said about creating AI services that customers can subscribe to.
With AI skills being hard to find and expensive, many organizations cannot take the first steps into AI. Even where AI engines are available under open-source models, the skills required to implement and train an AI engine to do what the organization requires present too many hurdles to overcome. On top of this, maintaining the skills of an AI specialist is not only difficult but if done correctly, it makes the AI specialist more of a target for competitors with better job offers. Many will end up working as AI engine developers or at an AI-specific vendor.
Leveraging MSPs for AI integration
This is where an MSP can help. Sure, even for MSPs, the costs of acquiring and maintaining skills around the deep technical aspects of AI may be too much for them, but self-provisioning and managing an AI stack is possibly – well, actually probably – not the right way to go about things.
There are AI services available in the same way that MSPs make their services available. Customers can subscribe to AI engines from various hyperscaler providers, such as Azure, AWS, and GCP. To end users, this may still not be what they are looking for. But, it may meet the needs of developers who want to try out AI within the organization, without the need to pay for expensive physical equipment and resources on-premise. Yet, it remains a technical mountain to climb to integrate such AI into existing systems. The hyperscalers are integrating AI into many of their services. Again, this may not meet the actual requirements of end-user organizations.
Integrating third-party cloud-based AI engines
MSPs who can directly integrate with a third-party cloud-based AI engine into their existing environment will achieve the best of all worlds. This can create many benefits:
- Use the AI capabilities to help them optimize their services.
- Leverage AI in their service delivery, such as enhancing their reporting capabilities, without the need to ‘own’ an AI engine.
- Offer AI engine as a service for their customers.
Developing essential skills for AI integration
MSPs will still need to build up certain skills to make this happen, though. To understand how large language models (LLMs) work, one will need knowledge. Customers will also gain access to vertical-specific models. This will enable them to incorporate the nuances and company-specific details they seek into their queries. In many cases, these models will come from the customers themselves.
MSPs will need to ensure that they completely air-lock the data used from the overall LLM. A customer’s own data and therefore intellectual property may well be included in the vertical-specific data. The data may also include personally identifiable data (PID) that requires specific handling for legal reasons. MSPs must ensure complete coverage of areas such as data sovereignty and data permissions.
However, the area where MSPs will require the deepest skills will be integration. AI is not a case of just throwing a lot of data at an engine and waiting for a miracle. For those who have tried using AI engines, it soon becomes apparent that the engine requires help in coming up with responses that the user is looking for.
As such, ensuring that AI integrations work the way the MSP and the customer expect them to will be key. This will involve extensive testing and optimization by the MSP before the customer gets to use the system. Continuous tuning of the system will be necessary to maintain good levels of output.
Facilitating effective utilization of AI
When it comes to offering AI as a stand-alone service, MSPs must be equipped to help customers use the system’s capabilities. Textual interfaces between users and GenAI are common. Customers will only get value if they understand the subtle nuances such as how some simple queries may cause incorrect, or useless, responses.
For example, asking an AI system, “How do I get new customers?” is likely to give responses that are, at best, dull, and at worse, misleading. Changing this to something more like “Based on last quarter’s most profitable customers, where are we most likely to be able to identify similar customers over the coming quarter?” will elicit a much more accurate and informed response.
MSPs should build up skills in this area. Being able to help a customer present the right queries to an AI system will be perceived as a very strong business value add. It’s essential to see them not just as an MSP offering access to an AI engine.
AI will undoubtedly change how MSPs operate, both through internal use and in how MSPs offer AI to their customers. Based on the market dynamics, just offering AI may rapidly become table stakes. The MSP will need to ensure that what they offer can differentiate them from the crowd.
Photo: NicoElNino / Shutterstock
