Why and How IT Leaders Can Embrace the AI Revolution

By Ofer Feldman, Stampli

If you manage an IT department, you've probably fielded plenty of questions over the past year or two about which types of AI solutions your business can benefit from, and how your team can implement and support them.

And you might also feel more than a little overwhelmed. Many conversations about AI's impact on IT operations focus on how building and managing AI applications require a host of new resources and commitments on the part of IT — like specialized AI/ML development skills, mastery of data quality concepts, and the ability to manage novel privacy and security risks associated with AI applications.

I'm here to tell you — as CTO of a company that specializes in building AI solutions for finance — that integrating AI tools for business doesn't have to become a source of frustration and increased load for IT teams. There is some level of effort required, to be sure, just as deploying any new solution requires some amount of work. But when IT departments approach AI strategically, rolling out AI solutions that bring real value to users doesn't have to be any more difficult than supporting more traditional types of software.

How AI Impacts IT

AI software certainly has some consequences for IT departments. There may be some new types of workflows to manage, new user requests to support, and new application deployments to track.

Related:AI Assistants: Picking the Right Copilot

But unless your business is actually building complex AI solutions from scratch — which it probably isn't or shouldn't because sophisticated, mature AI tools and services are available from external vendors, complete with support plans and SLAs — implementing AI is not actually that challenging.

That's because most third-party AI solutions boil down to SaaS apps that work just like any other SaaS: The vendor builds, manages, and supports them, with few resources and little effort necessary on the part of customers' IT departments. From the perspective of IT, implementing AI isn't all that different from implementing any other type of software.

The Varying Types of AI Solutions

It's important to note that the accuracy of the statements I just made can vary depending on which types of AI solutions your IT department is asked to implement and support. Some AI tools and services pose more challenges than others in various areas.

1. Data privacy and security

For instance, if you're deploying an AI-powered chatbot — even if it's developed by an outside vendor — your IT department will need to prepare to address challenges like the risk that the chatbot could reveal sensitive information that shouldn't be available to certain users, or that it generate responses that simply don't make sense. In that case, it would be important to work with the vendor to identify and erect guardrails that mitigate the risk of behavior like this.

Related:AI Quiz 2024: Test Your AI Knowledge

But in other cases, the risk profile of novel AI solutions is fundamentally lower. For instance, my company's primary product is an AI tool that learns customers' invoice coding practices and codes invoices automatically. For IT, there are really not any novel data privacy or security risks at stake here. The app ingests financial data, but so do plenty of non-AI applications. IT's responsibility when it comes to managing data security for this type of app boils down to vetting the vendor by reviewing its data management and compliance practices. The fact that the app uses AI doesn't change this process.

2. Data management

Likewise, the way IT departments manage data can vary depending on how their company uses AI. If you want to build a large language model (LLM) from scratch, or train a foundation model on a large and diverse corpus of your company's data, your IT team will need to invest a lot of time and effort in collecting and preparing the necessary data. That's a big lift.

But if the business opts for an AI-powered finance tool, for example, there is no need for complex data management on the part of IT. Apart from ensuring that the tool can access incoming vendor invoices, IT doesn't have to assume new data management responsibilities.

3. End-user support

Support requests fall into a similar boat. Complicated AI solutions that a business develops and manages itself may require the IT team to be able to troubleshoot complex issues, necessitating special AI-centric skills on the part of IT personnel. But with a SaaS AI app, IT can outsource this work to the vendor, with the result that the app has minimal impact on end-user support operations for the customer.

Learning to Love AI

I mention all of this to encourage my fellow CTOs and IT professionals to embrace their organizations' adoption of AI. Rather than viewing novel AI solutions as an added burden that the IT department must bear, they should take peace in knowing that the capabilities and resources that the typical IT organization already has in place will suffice for allowing their end users to take full advantage of AI — especially when businesses choose AI solutions that create value for users without requiring extra effort on the part of IT.

About the author:

Ofer Feldman is the co-founder and CTO of Stampli. He is a highly experienced software development leader specializing in developing complicated software solutions by leading, guiding, and motivating large, agile teams of developers.