The AI draft is the easy part. The source check is the job
In corporate communications, the first useful thing an AI assistant does is not write a finished announcement. It gives you a place to begin. It can turn a long technical brief into an organised list of claims, flag missing stakeholders, or reconstruct a scattered email chain into a timeline that people can check. That speed […] The post The AI draft is the easy part. The source check is the job…
The initial usefulness of an AI assistant in corporate communications lies not in generating a finished announcement, but in providing a starting point. It can convert a lengthy technical brief into an organized list of assertions, highlight missing stakeholders, or transform a disorganized email chain into a timeline for verification.
While the efficiency gained is valuable, the accuracy of the summary remains crucial. A smooth-sounding summary may misrepresent an exploratory conversation as an approved contract, an old quote as current, or a potential roadmap as a firm commitment. Generative models should be used as drafting and triage tools, not as the authoritative source of information.
Before a sentence reaches the final edit, it is essential to confirm three key aspects: what precisely is being claimed, what source supports it, and its legal and commercial status. Instead of immediately asking an AI to pen an announcement, it is more effective to leverage it as a parser. The model can break down each assertion into a table, listing the claim, its supporting document, the author or owner of the document, and whether written clearance exists.
This approach shifts the focus from making the content sound exciting to demonstrating what can be proven. While an AI-generated draft can speed up the process, it cannot replace the need for a human to verify the accuracy, approval status, and appropriateness of the claims for public release. Polished drafts can instill a sense of urgency to send, so having a clear list of pending checks can help in pausing and reviewing.
When using verbs like "launched," "partnered," "selected," "acquired," and "secured," it is important to understand their distinct meanings. An AI-assisted draft might draw from genuine historical correspondence and inaccurately portray it as a current partnership. To avoid this, one must refer back to the primary correspondence, verify the relevant agreement and approvals, and eliminate unsupported assertions.
This discipline applies to the AI company's communications as well. For instance, if the HKMA's published list names the company among the technology partners for the first GenA.I. Sandbox++ cohort, but the SFC's published annex does not reflect the same, there is a discrepancy. Similarly, if the company's CTO is listed as a contributor to OWASP's Agentic Skills Top 10 (AST10), an inconsistency should be rectified to match the public record.
As the industry moves towards more capable AI agents that can plan and utilize tools across multiple steps, it becomes increasingly challenging to establish permissions, oversight, and accountability. An agent with access to company data should not be automatically considered safe just because its output appears plausible. Implementing a "maker-checker" approach can help in this regard.
One component gathers information or prepares a draft, while a separate control process ensures the action adheres to defined rules and routes higher-risk decisions to a human. This two-step process ensures that AI is used effectively without compromising safety and accountability. The level of review should correlate with the risk associated with the communication.
An internal recap email, for example, does not require the same scrutiny as a public statement, a regulatory filing, or an earnings disclosure. However, each communication process should clearly indicate which facts are verified, who approved them, and what remains uncertain. The saved time due to AI-assisted communications should be utilized to trace claims back to their sources, consult subject-matter experts about the actual authorizations of a document, and verify whether the draft reflects the actual decision made, not just the one that was hoped for.
The goal is not a draft that sounds the most certain, but one whose confidence level accurately reflects the evidence supporting it. In communications, the source check is not merely a final proofreading step; it is an integral part of the work.
Written by urgent.news from e27's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.