Hardware asset management has never been the glamorous end of IT. It is receiving docks, serial numbers, warranty dates, stockrooms and disposal certificates. That is exactly why the AI promises in this space deserve a careful look. The work is repetitive and text-heavy, which suits Gen AI, and it is full of small decisions that agents could in principle take, which is why every vendor now talks about autonomous asset lifecycles.
We reviewed the announcements, documentation and release notes of the main HAM and endpoint tools in October 2026 and sorted what exists from what is promised. The short version is that natural-language assistants are real and mostly read-only, document extraction and lifecycle estimation have shipped in a few products, and agents that actually order, receive and retire hardware are early access or announced. This article lays out the tools, the use cases, the evidence on benefits and the conditions that decide whether any of it works. If your first question is whether your records are good enough to build on, start with our piece on how a HAM and CMDB maturity assessment helps.
What AI can and cannot do for hardware assets
It helps to be precise about the layers. Discovery tools find devices on the network and in management consoles. Reconciliation matches those findings to purchase, contract and ownership records. The asset record is the result. AI sits on top of that record. It can answer questions about it, summarize it, extract data into it from documents, estimate fields that are missing, and propose actions based on it. It cannot invent a device the discovery layer never saw, and it cannot repair a record whose owner field has been wrong for three years. That is why vendors in this space, from Lansweeper to Axonius, keep saying the same thing: the AI is only as good as the asset truth underneath.
It also matters whether an AI feature reads or writes. A read-only assistant that answers “which laptops go out of warranty next quarter” carries little risk. An agent that disposes of a device carries real risk. Most of what exists today is on the read-only side, and that is sensible.
Tool by tool: what exists today
As in our ITSM review, we label each item. Available means the source says it is live. Preview covers beta, early access and controlled release. Announced means we found an announcement but no availability statement. Several vendors give no label at all.
ServiceNow Hardware Asset Management
ServiceNow has the most shipped AI in HAM. According to its Australia release notes for HAM, Now Assist has two skills: one extracts asset data from attached documents and creates purchase orders, and the other generates asset insights that summarize lifecycle history, chain of custody, audit status and data gaps for both people and AI agents. The release also added approximated lifecycle dates, which estimate end-of-life and end-of-support dates when the manufacturer has not published them, plus real-time audits and resale of assets at disposal. A store listing describes AI Agents for ITAM as a team of agents shared across software and hardware asset management, without naming the individual agents or a release date. At Knowledge 2026 ServiceNow also said an asset lifecycle specialist was expected in June 2026. We could not verify that it shipped.
The ServiceNow product page claims 50 percent fewer manual tasks, 50 percent shorter inventory audits per facility and 42 percent fewer hardware assets. There are no customers or baselines behind those numbers, so treat them as marketing.
Freshservice and Device42
Freshworks made its reimagined IT asset management generally available for new customers on 31 March 2026, with Device42 discovery and dependency mapping built in. Existing customers migrate in phases through 2026 and early 2027. The announced Freddy AI Agent Studio and MCP Gateway can use asset context, but we found no asset-specific agents named.
Ivanti Neurons for ITAM
Ivanti’s AI self-service agent has been in controlled release since April 2026 and must be requested. Ivanti says it checks ITAM, CMDB and discovery data, such as ownership and permitted software, before it acts. Neither Ivanti source we found describes an ITAM-specific AI feature, so the benefit shows up in service requests rather than in asset management itself.
Lansweeper
Lansweeper’s spring 2026 update is a good illustration of the read-only pattern. Its MCP server, which gives outside AI tools access to assets, software, vulnerabilities and users over OAuth 2.0, is read-only and in beta. Lens Chat for natural-language questions and the Traffic Sensor are open beta. Flow Builder is generally available. A plugin for hardware refresh planning and coverage checks is listed without a stage.
Axonius
In July 2026 Axonius launched an AI Agent in preview and an MCP server in early access that translates natural language into its query language. A CMDB reconciliation workspace was announced without an availability date.
Oomnitza
Oomnitza’s May 2026 release notes describe an MCP server in early access, exposing asset data, search and workflows, with copilots able to trigger lifecycle actions. Access is by request through a customer success manager. It is one of the few sources that explicitly mention an agent doing something rather than only answering.
Flexera One
Flexera’s September 2026 update lists Hardware Lifecycle Intelligence in IT Visibility, backed by its Technopedia library, plus a sync with Jira Service Management Assets. The post mentions no agents or MCP. On 6 October 2026 Flexera announced a collaboration with Atlassian, with no launch date.
Atlassian Assets in Jira Service Management
At Team ’26 Europe on 7 October 2026 Atlassian announced a hardware asset management module for JSM with AI insights in the console, such as warranty expirations and device shortages, that can trigger replenishment workflows. The source labels it neither generally available nor beta. Discovery comes from Atlassian, Flexera or Lansweeper.
Microsoft Intune
Copilot in Intune offers natural-language exploration of device data, KQL generation for device queries and device summaries. It is included with Security Copilot and billed in security compute units. The more instructive story is the Device Offboarding Agent. It required admin approval before disabling devices, new setup ended on 30 April 2026 and Microsoft removed it on 1 June 2026. Other Intune agents are reported to be in public preview. If you build a process around a vendor agent, plan for the possibility that it changes or disappears within months.
Jamf
Jamf’s AI Assistant, described in June 2026, uses retrieval-augmented generation, is disabled by default and cannot change configuration. Write actions are planned and will require admin approval. At JNUC in September 2026 Jamf announced AI governance for Macs, with several items targeted for Q4 2026.
Certero, ManageEngine, Asset Panda and Snipe-IT
Certero’s CerteroX v9 AI Assistant answers plain-English questions over hardware, software and licensing data. It runs structured queries rather than free-form SQL and is off by default. ManageEngine’s Endpoint Central offers three predefined Zia agents, for device investigation, EDR alert triage and patch troubleshooting, restricted to admins and consuming Zoho credits. Asset Panda announced its Ursa AI platform in August 2026 with natural-language questions, photo-based inventory and compliance audits, and no availability details. Snipe-IT’s 8.8.0 release in September 2026 lists no AI features, though community-built MCP servers exist.
Where the evidence sits, use case by use case
The matrix maps seven common HAM AI use cases to the ten tools with the clearest documentation. As before, a blank cell means we found no evidence in the vendor sources we reviewed, not that the tool cannot do it.
The picture is consistent with what the vendors say in their own words. Questions and summaries are the most common feature. Lifecycle estimation and document extraction have shipped in a few products. Agents that act on the lifecycle are early access or announced almost everywhere, and the one agent we found that did something consequential, the Intune offboarding agent, was retired within months.
The use cases in detail
Asking questions of the asset record
Natural-language access is useful because most asset questions are one-off and awkward to build a report for: which devices are out of warranty in two quarters, which laptops have no owner, which models are still on a retired operating system. The quality depends on how the product turns a question into a query. Certero says its assistant runs structured queries rather than free-form SQL, and Axonius translates language into its own query language. That constraint is a good sign. Be wary of any assistant that generates arbitrary queries against production data without showing you the query.
Lifecycle forecasting and refresh planning
ServiceNow’s approximated lifecycle dates and Flexera’s Hardware Lifecycle Intelligence both address a real gap: manufacturers do not always publish end-of-life dates, and spreadsheets rarely track them. The urgent example is Windows 10. Support ended on 14 October 2025, and Omdia put commercial extended security update pricing at $61, $122 and $244 per device for years one to three, which makes replacement cheaper than extension for many devices by the second or third year. Planning that refresh without a reliable age and warranty record is guesswork. We cover the numbers in our maturity assessment article.
Extracting data from documents
Purchase orders, delivery notes and warranty certificates are where hardware data starts and where it is usually retyped. ServiceNow’s skill that creates purchase orders from attachments is the clearest shipped example. We found no verified product that extracts warranty or contract terms for hardware, so if a vendor claims it, ask for a demonstration on your documents.
Discovery, reconciliation and ghost assets
This is the use case people most want and the one with the least AI-specific evidence. Freshservice with Device42 offers continuous discovery, Lansweeper has a beta traffic sensor, and Axonius has announced a reconciliation workspace, but these are discovery and matching features rather than proof that AI finds ghost assets. We found no vendor source that demonstrates anomaly detection for missing or duplicated hardware.
Agents that order, receive and retire hardware
Oomnitza’s early access program and Atlassian’s announced replenishment workflows point in this direction. The shape that makes sense is a loop with clear gates, shown below.
Sustainability and disposal
ServiceNow’s asset resale at disposal is available, though it is not an AI feature. Flexera’s sustainability dashboard covers AI cloud spend, not hardware. For hardware, the practical value is simply a complete disposal record, which supports both data protection and the e-waste reporting that many organizations now owe.
The benefits, and what is actually evidenced
We found no independent, HAM-specific measurement of what AI features deliver. The numbers that exist are vendor claims, such as the ServiceNow figures above, or come from adjacent areas. CIO reported in October 2026 on customers using agents for IT work, including a West Monroe IT and HR agent that cut managed service cost by 40 percent, and a worked example from Glokal in which an agent’s headline saving of 60,000 euros shrank to about 36,000 euros once rework and checking were counted. That second story is the more useful one. Net the review time against the gross saving before you sign.
The independent evidence we do have is about the foundation. Flexera’s 2026 State of ITAM report, a vendor-run survey of more than 500 respondents, found that only 36 percent report complete visibility across their estate, down from 43 percent, and that confidence in on-premises hardware visibility slipped from 76 percent to 74 percent.
How it is evolving
From rules to agents, with a long stop at assistants. Early HAM tools used rules for matching and alerts. Machine learning arrived in normalization and lifecycle estimation. The current wave is Gen AI assistants, mostly read-only. Agents come next, and the vendors that have shown them have gated them: Oomnitza by request, Jamf with planned admin approval, Intune with a required approval that still did not save the product.
MCP is becoming the access layer. Lansweeper, Axonius, Oomnitza and Freshservice all offer or have announced MCP access, and community servers exist for Snipe-IT. That lets outside AI tools query your asset data directly. It also means your asset record becomes an input to agents you have not designed, so access control on the MCP server matters.
Vendor boundaries are moving. Atlassian is building HAM into JSM and partnering with Flexera for discovery data. Freshworks absorbed Device42. Expect more consolidation, and ask about data export.
The analyst view is cautious. Gartner said in 2025 that more than 40 percent of agentic AI projects will be canceled by the end of 2027, citing cost, unclear value and weak risk controls. The ITSM.tools survey from June 2026, which HCLSoftware sponsored, found 15 percent of respondents running partially autonomous agentic AI in ITSM and 6 percent largely autonomous, and listed poor data quality as the top barrier at 32 percent, ahead of governance concerns at 30 percent.
Prerequisites and risks
Data quality and discovery coverage. Flexera put it plainly: AI does not solve bad data. Lansweeper names completeness, consistency and currency as the three gaps. If your discovery misses remote and bring-your-own devices, the assistant will give confident answers about the devices it can see.
Human approval. Purchases and disposals need explicit approval steps. We found no vendor guidance on how to design them, which is why we show the loop above.
Status inflation. Many features in this review are beta, preview, early access or announced. Get availability in your edition in writing.
Cost and stability. Intune’s agents consume security compute units, ManageEngine’s consume Zoho credits, and the Intune offboarding agent was retired within months. Budget for usage and for change.
A readiness checklist before you pilot
- Do you have a named owner for the asset record and for each device class?
- Does discovery cover remote, bring-your-own and stockroom devices?
- Do purchase, warranty and contract data link to the asset record?
- Can you list which AI actions need approval, and who approves them?
- Is access to any MCP server limited and logged?
- Have you measured a baseline for audit time, refresh planning time and reconciliation effort?
Where a maturity assessment fits
Every vendor above says some version of the same thing: the AI needs trustworthy asset data. A maturity assessment is how you test that claim against your own estate before you pay for the feature. Start with the Hardware Asset Management maturity assessment, use the Asset and Configuration Management assessment if your CMDB is part of the picture, or see all assessment categories. If you are comparing products, our guide to hardware asset management and CMDB tools covers the platforms themselves.
Frequently asked questions
Can AI find ghost assets?
Discovery finds devices. AI can help match and flag inconsistencies, but we found no vendor evidence of AI-specific ghost asset detection. Start with discovery coverage and reconciliation.
Which HAM tool has the most shipped AI?
ServiceNow has the most documented shipped features, including document-based purchase orders, asset insights and lifecycle date estimates. Others are mostly assistants or are in beta or early access.
Are AI agents safe to let order or dispose of hardware?
Not without approval gates. Design the agent to propose and prepare, and have people approve spend and data destruction.
Do we need MCP?
Not to start. It becomes relevant when you want outside AI tools to query asset data. If you enable it, limit and log access.
Check your hardware asset data before you add AI
Take the Desqcon online maturity assessment for hardware asset management and see how ready your records, processes and discovery are. Or talk to a senior consultant about a deep-dive.
Take the HAM Maturity AssessmentChoose an Assessment CategoryTalk to a Senior Consultant
Sources and further reading
- ServiceNow community, What’s new in the Australia release for Hardware Asset Management, May 2026, and the ServiceNow HAM product page
- Freshworks, Freshservice IT asset management 2026 announcement: freshworks.com/theworks
- Ivanti press releases, 27 January and 21 April 2026
- Lansweeper docs for the MCP server, and Lansweeper Connect Spring 2026 blog, 19 June 2026
- Axonius press release, July 2026, on AI Agent and MCP Server
- Oomnitza release notes, May 2026
- Flexera, What’s New September 2026 and State of ITAM 2026: flexera.com
- Atlassian blog, Team ’26 Europe service collection, 7 October 2026, and diginomica, 12 May 2026
- Microsoft Learn, Copilot in Intune, including the Device Offboarding Agent retirement notice
- Jamf blog, 9 June 2026, and Jamf JNUC announcements, September 2026
- Certero, CerteroX v9 AI Assistant, June 2026; Asset Panda press release, 6 August 2026
- Snipe-IT release notes, version 8.8.0: github.com/grokability/snipe-it
- CIO, Where AI agents are showing real IT savings, 1 October 2026
- Omdia, Microsoft has ended support for Windows 10, October 2025
- Gartner press release, 25 June 2025; ITSM.tools State of Agentic AI in ITSM 2026
