|
We’ve been hearing for years about the theoretical threat AI poses to enterprise software apps like Salesforce, ServiceNow and Workday. While a disaster scenario for these firms hasn’t happened because AI generally isn’t reliable enough to duplicate what they do, some customers are still trying to use the latest AI models to replace those apps to save money. For instance, some customers of Databricks, which sells a data lake, a type of database for storing different types of corporate data, are developing AI agents that automate some IT help desk tasks, such as responding to employees’ requests for technical help and automating software maintenance, according to a current Databricks sales manager. This software industry segment is known as IT Service Management, or ITSM. The Databricks sales manager also sees customers developing agents to manage customer service and human resources functions, such tasks related to recruiting new employees. These AI projects are designed to handle parts of existing enterprise apps and aren’t yet full-blown replacements, this person said. Another example is Lakeside Software, which sells software that helps IT departments continually monitor the applications and devices their companies use to track the apps’ performance and spot potential problems early. Lakeside has its own data lake for customers to store data its software collects, and some customers are experimenting with AI that can use that data to automate IT helpdesk functions, said Tal Klein, the company’s chief marketing officer, on the sidelines of Microsoft’s Ignite conference in San Francisco. In at least one instance, a Lakeside customer went even further by using software from UiPath, a provider of task automation software, to train an AI model to mimic how its employees used HR software from Workday, Klein said. That customer, which he declined to name, has now moved off Workday and is using the bespoke AI model to handle HR functions like assisting new employees, he said. Of course, it’s one thing to hear from enterprise vendors about customers pulling off such feats than it is to hear from the customers themselves! Some companies we talked to that previously said they would do it, like a midsize software firm we quoted here, still haven’t done so. Despite these app replacement projects, it still isn’t clear if AI will become a well-worn escape route for companies looking to save money that way. The knock on the AI approach is that companies will still need to do the heavy lifting of developing the strong security and data management controls that underpin most enterprise apps. “Getting into ITSM isn't as simple as putting an [AI model] on top of a data lake; you need the workflow integrations that take action,” a ServiceNow spokesperson said in a statement via email. Workflow integrations are custom-designed software that lets ServiceNow’s products connect with those from other providers in order to exchange data or update a database. General Catalyst, Anthropic Bet on Public Sector Customers By now it’s clear that many AI buyers need help from consultants to make the technology work. Those buyers include government agencies, such as those in the state of Maryland. The state recently awarded IT consulting startup Percepta and Anthropic—which has been increasingly providing hands-on assistance to customers using its technology—a contract to overhaul how the state responds to public requests for real estate records. Together, the companies developed a chatbot that helps the Department of Environment find and disclose documents residents requested in less than a day. That process, previously handled by human employees, would take weeks per request, according to a state official who spoke to me about it. Separately, Anthropic engineers developed a new chatbot to answer Maryland residents’ questions about welfare benefits programs like food stamps and medicare so they can sign up for the programs, the official said. The two chatbots have been used by more than 600,000 people since they launched in June, the official said. To pay for the chatbots, Maryland tapped a Rockefeller Foundation grant geared to AI programs, the official said. To develop the chatbots, engineers from both Anthropic and Percepta, which is wholly owned by venture firm General Catalyst, spent roughly five weeks working inside the Maryland agencies, according to Percepta CEO Hirsh Jain, who was formerly a Palantir executive. While many large organizations have struggled to get AI pilot programs off the ground, Percepta is betting there will be demand for hands-on consulting services to make AI work, especially among public sector customers that may not have many AI engineers on staff. “A huge part of how we operate is by embedding directly with our customers on the ground working hand in hand with them,” Jain said. “The idea is being able to lean on innovation to make government services run more efficiently.”
|
