Answer capsule
ALEX's support guide says new conversations do not automatically remember prior discussions, while information stored in Preferences persists across conversations. A buyer should test how a changed role, team, goal, or communication preference is reviewed, corrected, and retired before stale context shapes later leadership guidance.
What the source establishes
- Admired Leadership's current support guide says every new ALEX conversation starts fresh and does not automatically remember previous conversations.
- The same guide says ALEX Preferences can provide persistent information that carries across all conversations.
- The guide says ALEX uses proprietary Admired Leadership content plus general knowledge through July 2025, does not search the internet, and may generate plausible but inaccurate cases or facts.
- The public guide does not establish preference-field structure, source and update dates, review prompts, conflict handling, deletion propagation, administrator visibility, or buyer-specific controls.
Inventory what persists and why
The direct answer is to create a preference inventory before asking leaders to rely on continuity. For each persistent field, record the value, purpose, source, person who entered it, date confirmed, sensitivity, affected coaching uses, permitted viewers, review date, and deletion condition. Separate durable presentation choices such as language or response style from facts that can change quickly, including title, reporting line, team membership, priorities, authority, health or accessibility needs, and an active workplace conflict. A useful preference should reduce repeated setup without silently converting a temporary situation or one person's interpretation into a standing fact across every future conversation.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Test fresh conversations against changed context
Use synthetic cases to compare a new conversation with no preference, a correct current preference, a stale preference, a conflicting statement in the new thread, and a preference that has just been removed. Change the leader's role, team, location, goal, communication style, and decision authority one at a time. Observe whether ALEX surfaces the persistent assumption, asks for confirmation when it matters, follows the current statement over stale context, and avoids carrying one scenario into an unrelated one. Record the preference version, new-thread input, response, correction, and reviewer disposition. Starting a conversation fresh is not operationally fresh if unseen persistent context continues to frame it.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Exercise correction, deletion, and fact verification
Correct an inaccurate preference, remove a sensitive one, and verify the change in a later conversation, search or history view, export, and any connected administrative record allowed by the buyer's configuration. Test cached sessions, multiple devices, voice and web access, and an interrupted update. Separately, ask time-sensitive questions that conflict with the support guide's July 2025 general-knowledge boundary. The system should not turn persistent user context into current external fact or present an invented case as evidence. Require links or authoritative verification outside the coaching system for laws, policies, people, prices, product behavior, and other facts that may have changed.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Set review and expiry before organizational use
Give leaders an accessible place to review and change preferences, explain what new conversations forget and what Preferences retain, and prompt reconfirmation after a role change or a defined interval. Set shorter expiry for sensitive or consequential facts and a clear route to correct or delete them. For an organizational deployment, define whether administrators can create, infer, view, export, or report on preferences and keep sponsor analytics separate from conversation and development content. Monitor stale-context corrections, hidden-assumption complaints, deletion failures, cross-device inconsistencies, and guidance affected by outdated facts. Expand only when persistence remains visible, user-controlled, bounded, and recoverable.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Limitations and unknowns
Admired Leadership is the provider source. Its current support guide describes ALEX's web and iPhone access, proprietary leadership content, document review, general knowledge through July 2025, no internet search, possible plausible factual errors, fresh-conversation behavior, persistent Preferences, language coverage, and positioning as a complement to human development. It does not independently establish preference fields, collection and inference, review and expiry, conflict resolution, deletion propagation, administrator access, model and version, consistency, privacy implementation, accessibility, coaching quality, leadership development, or organizational outcome. Current support, privacy, security, model, subscription, and enterprise terms, preference and data inventories, representative synthetic stale, correction, conflict, and deletion tests, user and sponsor boundaries, and qualified coaching, learning, people, accessibility, privacy, security, procurement, and legal review control.
Decision test
Ask whether the source changes the decision itself, the evidence required, the implementation sequence, or only the language used to describe an existing capability. Record which claims are directly supported, which are provider statements, which require an independent test, and which remain unknown. A source-linked review should make uncertainty easier to see, not bury it inside a blended score.
Questions to take into review
The publication supports research and executive decision preparation. It does not provide legal, financial, accounting, employment, clinical, cybersecurity, investment, procurement, or implementation advice.