An answer should be connected to the material that supports it
The documents need to be read together, subsequent changes identified, and termination on notice distinguished from termination for breach. The conditions for each option also need to be checked. Correspondence may raise further questions: was the breach notified, was an opportunity to remedy it given, and how did the other party respond?
Source-grounded AI connects analysis to specific documents and relevant legal materials, rather than relying solely on the model's general knowledge. In Lexandro, conversations can draw on uploaded documents, selected legal collections, and the web. Users select the materials appropriate to the task, while access to legal databases depends on the jurisdictions and permissions enabled for their account.
Citations accompanying supported claims allow users to inspect the underlying document or external page. This gives professional review a concrete starting point. Instead of asking, "Does this sound convincing?", the user can ask more precise questions:
• Does the cited passage actually support the conclusion?
• Has a later amendment changed that provision?
• Is there an exception in another document?
• Does the legal material apply to this jurisdiction and the relevant period?
• Which parts of the answer are established facts, and which are interpretations or assumptions?
A citation starts the review — it does not guarantee accuracy
Imagine that the original contract provides for one notice period, while a later amendment provides for another. An answer could accurately cite the original clause and still misstate the options for exiting the agreement. The problem would not be a fabricated citation, but an incomplete analysis of how the documents relate to one another.
A useful result should therefore distinguish three levels:
1. What the document says. A specific provision or recorded fact.
2. What it may mean. An interpretation in light of other documents, the circumstances, and applicable law.
3. What still needs to be confirmed. An unresolved question, a missing document, or a fact that could change the conclusion.
Connecting answers to documents makes review easier. It does not remove the need for legal judgment.
Sometimes the most useful finding is what is missing
In the supplier-replacement example, correspondence might refer to a notice of breach without including the notice itself or evidence of its delivery. The useful result is not a categorical statement that the conditions for termination have been met. It is a clear warning:
"Before relying on this option, confirm the contents of the notice, the date of delivery, and any period allowed for remedying the breach."
That finding directs the next step. The team knows what to obtain and why it matters to the decision. It is also important to distinguish "not found in the materials reviewed" from "does not exist." Similarly, the absence of a warning does not establish that the documentation is complete or that every risk has been identified.
How Lexandro connects work on a matter
Verifiable analysis depends on more than an individual answer. It also matters how documents, questions, and working instructions are organised.
Workspaces keep the matter together. Lexandro workspaces bring together documents, conversations, and saved background information about a matter. This supports continuity without requiring users to explain the basic circumstances again in every conversation. Saved background information is not, however, the same as documentary evidence. It can guide the analysis, but it is not itself a document that can be cited. A claim requiring proof should be distinguished from a working assumption or business objective.
Source selection defines the scope of research. For an analysis of a contractual relationship, the starting materials might include the contract, amendments, general terms and conditions, and correspondence. Where the question requires checking a legal rule, users can include appropriate available legal sources. For current public information, they can include the web. The goal is not to include as much material as possible, but to select material that fits the question. Research results should not be treated as a guarantee that every relevant statute, judgment, or item of evidence has been found.
Skills help teams reuse a methodology. Skills in Lexandro are reusable instructions. Teams can use them to standardise checklists, risk criteria, and output structures. For example, a review methodology can require the analysis to present established facts, available options, inconsistencies, and unresolved questions separately. Skills guide how the work is performed. They do not automatically select evidence, expand access to documents, or constitute autonomous agents in their own right.
A better request asks for more than a conclusion
Instead of simply asking, "Can we replace the supplier?", a more useful request would be:
"Assess the options for ending the existing contractual relationship so that we can replace the supplier. Read the contract, amendments, general terms and conditions, and relevant correspondence together. For each option, identify its basis, the conditions that must be met, the main risks, and what still needs to be confirmed. Clearly distinguish facts, interpretations, and assumptions, and cite the documents supporting key claims."
This defines a deliverable that can be professionally reviewed. It does not guarantee completeness or accuracy, but it sets clear expectations for the analysis.
The real value is an answer you can verify
Speed is useful. But a fast answer that conceals uncertainty can make decision-making harder.
For a legal team, a high-quality AI result is more than well-written text. It is an analysis that shows its supporting material, distinguishes facts from interpretations, and clearly identifies unresolved questions. Lexandro supports this way of working by connecting documents, selected legal materials, citations, workspaces, and reusable instructions through Skills.
The goal is not to eliminate professional review. It is to give that review a clearer, more useful foundation.