In June 2022, San Mateo County Sheriff’s Office organized a “day of fun” down at the Coyote Point Shooting Range. The department’s IT team and staff from C3 AI, a $2.4 billion market cap company founded by billionaire Tom Siebel, were invited to celebrate their work on a new surveillance system intended to unify police data from agencies across the county into a single app. After lunch, C3 AI staff could practice shooting targets at the same facility where the cops trained. The choice of venue was not accidental: In 2018, Siebel’s foundation gave $4.5 million to the county to fund the range’s refurbishment.
Working closely with C3 AI and its longtime partner, Amazon Web Services, the Sheriff’s Office wanted to connect masses of data from at least 15 different agencies under its jurisdiction, collected from devices like license plate readers, surveillance cameras, arrest records, 911 calls and historical police databases. At the core of Project Sherlock, named for Arthur Conan Doyle’s fictional detective, was artificial intelligence that could piece the data together, find leads for investigators and give situational awareness to cops on the beat. The ultimate aim was to dramatically speed up the grunt work of policing.
For C3 AI, Sherlock was set to be a flagship contract for the company’s local government division. And so crucial was it that Siebel himself was deeply involved in the early stages of the work, down to writing sections of the contract and hosting San Mateo cops at his offices in Redwood City.
But $12 million and three years later, Project Sherlock has struggled with usability issues, feature oversights and significant delays, in some cases well over a year, according to public records obtained by Forbes. A timeline of the planned Sherlock rollout, obtained via public records, showed that 14 agencies were due to go live with Sherlock by January this year at the latest. Seven of those agencies — Atherton, Belmont, East Palo Alto, Foster City, Menlo Park, San Bruno and South San Francisco — told Forbes they didn't have access to a fully functional product and therefore hadn't seen any investigational benefit.
“My no BS assessment is that we’ve been working with them for two years and they have a barely functional product.”
San Mateo County Sheriff's Office claimed that in fact the Foster City and San Bruno police departments did have access, along with Daly City, Redwood City and San Mateo. The agency didn't clarify, however, whether officers at onboarded agencies were actively using Sherlock or if they were like Foster City, which was due to go live as of May 31 last year, but, per chief Cory Call, was “not seeing any benefits yet as this project is still in the developmental stages.” Similarly, San Bruno’s former interim chief Susan Manheimer told Forbes in June the department was using a limited, beta version of Sherlock.
Menlo Park Police Department chief David Norris said all 14 agencies were “behind the implementation schedule” in the timeline. “While the technology itself is onboarded, the practical implementation has been delayed,” Norris said. Belmont lieutenant Clyde Hussey, his department’s project manager for Sherlock, confirmed the delays, adding that he was unable to explain them. “I wish I had a good answer for you.” The other agencies didn’t respond to requests for comment.
A source close to agencies in the county said there was a sense of disappointment among Sherlock's participating departments. "Those it has rolled out to are really underwhelmed and waiting for it to start to develop,” they told Forbes. The source said it was “crazy” the product was still in development after three full years, adding it hadn’t “caught fire” even in those departments using the beta version. “Almost all agencies I’ve spoken with are looking forward to this project ending so they can get a robust system that gives them real-time information.”
The stuttering progress of Project Sherlock, the name and nature of which is reported here for the first time, shows how even with the backing of a Silicon Valley billionaire in Siebel and one of the world’s most powerful companies in Amazon, AI’s much-hyped use in policing is running into bureaucratic stumbling blocks and problems performing as promised. In San Mateo County, the hope that AI plus mass collection of surveillance data would get vital information into the hands of cops faster has yet to be fully realized.
San Mateo County Sheriff’s Office head of communications Gretchen Spiker blamed delays on “efforts to make improvements prior to onboarding additional agencies.” She said Sherlock remains a “work in progress,” adding, “The Sheriff’s Office and partner agencies continue to work closely with C3 AI to make the necessary improvements to the system.” C3 AI’s director for state and local government projects Grant Guttschow blamed Sherlock’s delays on what he described as inherent bureaucratic challenges that come with deploying such a sprawling surveillance system to multiple agencies. “We have frequent meetings with all the participants and as you can imagine the bureaucracy in a county-wide collaboration can sometimes slow down project timelines,” he said.
Neither San Mateo County nor C3 AI were able to provide data to show how Sherlock had made a significant impact on policing in the region. Guttschow said that there had been multiple “success stories” coming out of the initiative but declined to provide specific cases where Sherlock had provided significant support to investigators.
The Sheriff’s Office provided a single example, where C3 AI was used by patrol deputies on a violent battery case. “The system played a key role in helping deputies access critical information, which resulted in investigative leads and a subsequent arrest,” said Spiker. She said there were several other examples but could not detail them because she didn’t want to jeopardize ongoing criminal proceedings.
Large-scale police tech projects can often be hit by delays and functionality failures, but AI projects in policing can also be undermined by marketing hype that brings about a fundamental misunderstanding of how the tech actually works, says Aaron Rosler, a former criminal investigator with the State Department’s diplomatic service. “Being successful in this space can be difficult to quantify. Some software can be successful at doing its programmed analytical goals but will fail because agencies implement it poorly,” Rosler, now an associate manager at consultancy Kroll, told Forbes.
Sherlock is among the many “real-time crime centers” or RTCCs now being marketed to law enforcement agencies. Public safety tech companies like Axon, Peregrine and more recently Flock all offer such systems, which take public records and disparate surveillance data and route it to rooms full of intel analysts, who use AI to mine it for investigative leads.
C3 AI was well positioned to be a big player in the RTCC game and Sherlock a bellwether contract. Siebel made his fortune in handling big data, selling customer relationship management software provider Siebel Systems to Oracle in 2006 for $5.8 billion. In 2009, he founded C3 AI, which offers AI models that can help government and private companies reap insights from massive datasets, and in 2020 took it public at a $10 billion valuation.
Siebel had another reason to be optimistic about taking a big slice of the $11 billion police tech market: His partnership with Amazon Web Services (AWS). In 2017, Siebel said C3 AI and AWS jointly sold “every account worldwide” and claimed to be on speed dial with Amazon CEO Andy Jassy, then the CEO of AWS. Since 2023, Amazon has provided funding to C3 AI to build out its law enforcement product, and the two companies have an agreement to integrate their technologies. In San Mateo, a Sheriff’s Office Sherlock lead said AWS funds for C3 AI had covered the entire $600,000 Daly City pilot for Project Sherlock. Aisha Johnson, AWS comms director, said it was up to partners to decide how best to use partnership funding. She declined further comment on this story.
By mid-2023, C3 AI appeared to be moving ahead at speed, celebrating an award from analyst firm IDC given to the Sheriff’s Office for data-driven policing. Yet at the time, records show the project was not widely used by San Mateo cops. That year, one senior Sheriff’s Office analyst emailed a cop at another agency with a candid assessment of C3 AI’s technology: “My no BS assessment is that we’ve been working with them for two years and they have a barely functional product. It does what it says it does but not well. Lots of upside potential but I’d wait until they have a finished product.”
Guttschow said that he couldn’t speak to that particular comment, but suggested any complaints about the tool were largely coming from “disgruntled” users. He said Sherlock had “matured significantly” since 2023, telling Forbes it was “making progress, and we are proud to support everyone in participating agencies,” adding that “many happy analysts use the application today.”
But some of those San Mateo County Sheriff’s Office officers who have gotten access to Sherlock have raised issues. In emails obtained by Forbes from this March, a captain wrote to C3 AI asking why they couldn’t see a criminal history of a suspect they were interested in, when they knew the individual had one. The whole premise of Sherlock was to ensure quick and easy access to data on potential suspects, but the captain was told by a C3 AI employee that it wasn’t pulling any data from San Mateo County Sheriff’s records and case management system that was created prior to 2012. “Are we ever going to pull all the reports in?” the captain asked in one email, and was told by a C3 AI staffer that it was not scheduled. Sheriff’s Office spokesperson Spiker said the issue arose because the Sheriff’s Office transitioned to a different records management system in 2012 and Sherlock is only integrated with the current one.
There were more concerning errors. Earlier this year, a Sheriff’s Office employee warned that C3 AI’s software was ingesting “restricted cases,” which were sensitive in nature and should only have been visible to select individuals. There were also repeat cases of downtime and agents unable to login due to bugs. Guttschow said the company was always on hand to improve the app and iron out any flaws.
Siebel’s plan had been to take the Project Sherlock model and offer the same to other agencies. But C3 AI has lost ground to rivals. Guttschow claimed C3 AI had 20 customers using its law enforcement-focused products (he declined to name any or say if that number includes the 16 San Mateo departments). By comparison, Axon says its rival Fusus product is used by over 250 police departments across America. Peregrine also claims to have more than 250 public safety contracts and has seen its private valuation equal C3’s market capitalization at $2.5 billion — even though Siebel’s company has a much broader remit than local policing and an eight-year headstart on Peregrine.
C3 AI’s broader business is also struggling. After recording two years of consistent revenue growth, it dropped 35% from quarter-to-quarter in the most recent results, when it posted $70 million in revenue and a $125 million loss. Siebel described the results as “completely unacceptable.” The founder, who has been suffering from health and eyesight issues related to an autoimmune disease, announced in July he was stepping down as CEO after 14 years at the helm, though he’ll stay on as executive chairman. “I am confident the company is positioned to accelerate going forward,” he said at the time.
Earlier this year, according to the email records, Siebel’s law enforcement tech team was continuing to sell the promise of their software to Placer County Sheriff’s Office, in California. “They are very interested in deploying a similar application to Sherlock,” the company’s AI engagement director wrote to the San Mateo Sheriff’s Office staff, asking for assistance for a demo.
It ended in disappointment. “C3 AI did reach out to us,” Placer County’s communications manager Elise Soviar told Forbes. “We opted not to use their product.”