During change, people often need answers at the exact moment they are trying to act: “What changes for my role?”, “Where is the new process?”, “What do I do if the system fails?”, or “Who approves this now?” A chatbot can reduce friction by turning approved change information into a real-time support channel. In change-management terms, it is an asset, support mechanism, and communication design intervention that helps people move from awareness to knowledge, ability, and reinforcement [Prosci ADKAR]. (prosci.com)

The chatbot should not replace sponsors, people managers, training, or employee listening. Prosci’s benchmarking emphasizes active sponsorship, front-line engagement, frequent open communication, and manager support; it also reports that employees often prefer supervisors for personal-impact messages and senior leaders for business-rationale messages [Prosci, 2025]. (prosci.com) The chatbot’s job is to make consistent answers easier to find, capture emerging questions, and route complex issues to humans.

A change-support chatbot is a digital channel, usually embedded in Teams, Slack, an intranet, a help portal, or a project site. It answers approved questions, links to job aids, explains next steps, and escalates unanswered or sensitive issues. It may be rule-based, search-based, or AI-enabled.

Use it when the change affects many people, has repeated questions, involves role-specific steps, or has a compressed go-live period. It is especially useful for system implementations, process redesigns, policy changes, relocations, shared-service transitions, and phased rollouts. Avoid using it as the only support channel for high-anxiety changes such as restructures, job impacts, or safety-critical changes. Employee voice requires complementary channels and leadership/manager follow-through, not a single mechanism [CIPD Employee Voice]. (cipd.org)

Start four to eight weeks before launch, earlier for regulated or AI-enabled uses. Inputs should include the change story, impact assessment, role map, training plan, process documentation, approved FAQs, known resistance points, support model, and escalation paths. GOV.UK recommends estimating likely enquiry types, channel volumes, response times, and service levels when setting up support [GOV.UK Service Manual, 2016a]. (gov.uk)

Decide who owns content, who approves answers, what the bot may not answer, when it must hand off to a person, how analytics will be reviewed, and how privacy/security/accessibility will be handled. If the bot uses AI, apply risk governance: NIST’s AI RMF frames AI risk work around governing, mapping, measuring, and managing risks [NIST, 2023]. (tsapps.nist.gov) Use plain language so employees can understand the answer and complete the task the first time [Digital.gov]. (digital.gov)

A company is rolling out a new procurement system to 2,000 employees. The change team launches a Teams chatbot two weeks before go-live. An employee can ask, “How do I raise a purchase request after July 1?” and receive the role-specific steps, a link to the job aid, and a button to open the training module. If someone asks whether their approval authority is changing, the bot provides the approved policy and routes exceptions to the procurement support queue. If someone asks about job security, the bot does not improvise; it directs the employee to their manager and HR contact.

The change manager reviews the top unanswered questions each morning. When many users ask about cost-center errors, the team updates the job aid, asks managers to reinforce the fix in huddles, and adds a chatbot answer. The chatbot becomes both a support channel and an early-warning system.

Track chatbot usage, repeat users, top questions, unanswered questions, escalation volume, time to answer, feedback rating, and content update cycle time. Pair these with adoption indicators such as training completion, correct process use, reduction in errors, fewer duplicate help-desk tickets, manager confidence, and employee sentiment. For AI-enabled bots, also monitor reliability, inappropriate responses, privacy/security incidents, and handoff performance [Microsoft Research, 2018; NIST, 2023].