Artificial Intelligence (AI) is an unprecedented technology that many enterprises are racing to keep up with.

And, while many businesses are beginning to use AI and expect to ramp that up in short order, they continue to struggle because they can’t access all their data and connect it to advanced AI systems, according to the ninth annual Connectivity Benchmark Report from Salesforce Mulesoft.

Notably, 62% of respondents say their organization is not equipped to unify data systems to benefit from AI. Also, 75% are struggling to integrate data insights into user experiences.

But these challenges must be overcome, said Param Kahlon, EVP and GM at Salesforce, as “AI is a once in a lifetime Application. It’s going to have a fundamental impact on how businesses operate. It’s going to impact every business, small or large.”

More LLMs, but trepidation and data issues, too In its survey, Salesforce polled 1,050 IT decision makers across the globe, finding that the vast majority (80%) are using predictive and genAI models and expect a 69% increase in the average number of large language models (LLMs) they’ll use over the next three years.

However, while 85% of respondents say that AI has ultimately increased developer productivity, there are still significant barriers to adoption, including integration issues (reported by 90% of respondents) and security (according to 79%). There are also strong concerns around AI bias and ethical use and the technology’s impact on carbon footprints.

This dilemma is directly in line with digital transformation woes: Nearly every leader polled (98%) said they were dealing with significant challenges in that area, the biggest being data silos (81%) and tightly-coupled legacy systems that are highly dependent on one another (72%).

Kahlon said that there are an average of 991 apps in an Enterprise — yet just over a quarter of those are connected.

“Enterprises are using hundreds of systems to run their business,” he said. “Typically, there aren’t great connections across those systems. Data is living in silos.”

But AI thrives on data — and lots of it — and models must have access to it to provide quality, relevant outputs. Most importantly, though, enterprises must be able to make decisions about those outputs.

“AI can make a lot of predictions,” said Kahlon. But it all comes down to an organization’s capabilities, as they can be “bottlenecked by human capacity.”

If a decision can’t be acted on, “that prediction is not worth anything,” he said.

Automation a productivity partner Enterprises are putting more and more pressure on their IT teams — the number of projects they are being asked to tackle rose by an estimated 39% just last year alone, Salesforce found — and they are increasingly relying on automation.

Specifically, one in three teams is turning to robotic process automation (RPA), a rise from 13% in 2021 to 31% in 2023, according to the survey.

Developers gain nearly 2 hours per week when they use automation, and enterprises say it also increases operational efficiency, decreases costs, supports scalability and reduces the number of support tickets.

By far, productivity is a huge benefit, Kahlon said. AI can work side-by-side with developers to take over “mundane, simple types of tasks.”

“GenAI really creates this experience of co-developing,” he said. “A developer does not have to start from scratch when writing a piece of code.”

For instance, they can prompt sample code, and the AI will provide context and programming language syntax. Developers can then make tweaks and “adjust for what they really need,” said Kahlon.

What used to take hours can be done with a “back and forth” with genAI, he said.

“Your job really then becomes that of an expert,” said Kahlon. “The productivity boost developers will get with genAI is really immense.”

Similarly, genAI can lower the barrier for non-technical workers not as versed in writing code, he said.

Naturally, AI will be able to tackle more tasks as it gets better and error rates decrease. These abilities will not only be more in number, but “more complex types of tasks that are not always a straight line execution,” said Kahlon.

AI is revolutionary — but tread carefully GenAI is a maturing technology that shows immense potential. To get the most out of it, enterprises must develop a strategy — now. You can’t sit back and watch it mature, Kahlon said.

“Enterprises can’t afford a wait and see approach,” he said. “Unlike other technological developments, this one’s going to move really really fast. It’s not going to take years for adoption, it’s going to take months, if not weeks. Everybody should start now.”

Notably, have an approach to data and have a “holistic set” of available data, he advised.

Furthermore, “the domain of what is available from public large language models is amazing,” said Kahlon. Enterprises should adopt these models and differentiate themselves by applying proprietary data to tweak suggestions into their business context.

Most importantly, focus on trust, Kahlon emphasized. “You need to be making sure that you don’t break the values that you stand for.”

Don’t put customer data at risk, he said; ensure that you have an approach that allows you to adapt and accommodate immaturities in gen AI (such as hallucinations).

Ultimately, “make sure that you pursue this carefully,” Kahlon said.