The interface can move. The dependency can stay.
November's governed AI integrations keep operational permissions in underlying systems while assistant interfaces compete for the customer relationship.

Author and editor: Mike Ye. AI research analyst: Ella.
An AI assistant can become the place a customer starts work while an existing software provider remains the place that work must be authorized and completed. That separation is the most useful investment question emerging from Strategic Signal's November evidence.
Our judgment: in these integrations, operational Alignment remains with the underlying system's permission layer, while Interface is contested. All 27 November records that permit action execution keep it conditional on those permissions. LeanData and Impartner provide concrete examples of choice across independently operated assistants, weakening any one assistant's leverage. Separately, 21 of the 44 records list more than one AI recipient, although that count includes products from the same provider and is not a count of independent alternatives. We would change this view if an assistant could write directly to the relevant records without the underlying permission layer, or if a provider reported lost paid seats tied to assistant use.
Read Interface and Alignment together
Within the Four Forces of AI Power, this is an Interface question—who controls the entry point to the customer—and an Alignment question—who grants and constrains access and action. November's records do not establish a change in Compute or Energy power.
The map separates who grants and can withdraw permission from who controls the interface. Its OpenAI and Anthropic pages locate the assistant providers within those dependencies. LeanData, ThoughtSpot and Infor do not yet have mapped company pages in the checked AI Lattice v0.5.9 company table; their case-level evidence here comes from Strategic Signal.
LeanData's connection makes scheduling and orchestration available through Claude and ChatGPT while retaining existing rules. ThoughtSpot supplies governed business context to Claude and accepts outputs back into its own analytics environment. Infor permits selected customers to initiate user-approved manufacturing changes through Claude. These are observed access structures, not proof of earnings gains.
Supporting Strategic Signal records: SS-AI-2026-F6948782, SS-AI-2026-9FBB5BAF, SS-AI-2026-BDD980ED, preserved in archive 000045. Primary announcements: LeanData, ThoughtSpot, Infor.
A decision rule for evaluating software exposure
Ask what the customer would still need if most interaction moved into an assistant.
| Role the provider retains | What to investigate |
|---|---|
| Source of licensed information | Whether equivalent information is obtainable elsewhere on acceptable terms |
| Interpreter of business context | Whether its definitions and accumulated configurations are difficult to reproduce |
| Controller of operational changes | Whether its permissions, validations and records remain necessary to complete work |
| Commercial relationship | Whether it can charge for usage or outcomes as interface activity changes |
This is a diligence sequence, not a new scoring framework.
The distinction is economically important. A provider with useful information can face substitution. A provider that must validate and record a consequential change may retain a stronger negotiating position. That is our hypothesis about the mechanism, not a demonstrated valuation premium for the companies named.
The assistant also has a potential advantage: it can direct the task to a provider, aggregate alternatives or become the customer's default place to work. Its position strengthens if that control converts into repeat use and monetization. The records do not yet quantify that conversion.
What would change our judgment
The thesis would gain support if customer evidence showed that assistant-mediated activity increased paid use of the underlying system, preserved renewals or created an additional commercial channel.
It would weaken if comparable workflows could bypass the provider, customers removed paid licenses without losing useful capability, or contract disclosures showed that the assistant captured the value while the provider's economics deteriorated. Those are tests to investigate, not claims about present results.
More connectors alone would not settle it. The November cohort contains 44 reviewed announcements, including 33 connector arrangements and 27 with action execution. Those overlapping counts describe the selected registry sample. They do not measure adoption, revenue growth or bargaining power.
The implication for capital allocation
For acquisition diligence and investment research, identify the indispensable step before assigning value to the AI integration. Then examine who owns that step, who can revoke access, whether alternatives exist and what customers actually pay.
An assistant integration could improve a provider's distribution, expose it to substitution, or do both. The relevant evidence is the relationship between retained control and realized economics.
Strategic Signal holds the dated arrangements and forecast follow-up; AI Lattice maps dependencies; exmxc develops the judgment. No new company rating, sPEG call or Index score is issued by this brief.
Evidence: Strategic Signal November month-to-date outlook, evidence through November 19, 2026; frozen archive 000045. Shared Mike Ye authorship means the two publications are related analysis, not independent corroboration.
Published by exmxc. Not investment advice.