Are Enterprise AI Systems Helping Or Hurting The Corporate Comms Function?

As corporations accelerate their AI transformation plans, the central question enterprise AI systems are trying to answer is how much time can be saved. But AI systems should not be designed by tech teams alone. The teams affected should have a say in how it is designed.

That's what Lee Jie Qi, senior VP of Edelman Intelligence said at a panel session held at the PRovoke Media Asia-Pacific Summit. The discussion centred around the gap between corporate AI systems and their teams’ readiness to deploy it.

Another panelist, Jess Tang, founder of mavic.ai, said this became clear to her as a communications practitioner who transitioned into building an AI platforms.

“They see a text input and a text output with the correct word count, or a picture that has been created,” she said. “I look at it and say, ‘This is horrible. I cannot accept this.’ Technically, it works, but it is not a piece of communication.”

I completely agree with this. Because I am also a communications practitioner who builds AI content writing agents.

Technical AI Success Is Not The Same As Communications Success

AI technical teams know how to design AI systems, but comms professionals know what good communications look like, and more importantly, what it takes to produce good communication strategies and assets. The risk of an AI system built FOR corporate communications professionals and not BY them is not the risk of an AI system malfunction. Comms professionals can adapt back to a life before AI.

The point is not that AI is unreliable. The point is that reliability in one domain is not a guarantee of reliability in another. The controls most organisations have built around AI-mediated communication tend to be technical and compliance-oriented. Does the output contain mistakes or misinformation? Does it violate policy? Does it meet legal standards?

Technical accuracy — correct grammar, coherent structure, factual correctness — is different from communications quality. A message can be technically flawless and still miss the mark.

Communications quality depends on context: the specific relationship between the organisation and its audience, the cultural environment in which the message will land, the strategic narrative the organisation is trying to advance, and the reputational risks that are particular to this sector, this market, this moment. These are not variables that can be encoded into a general-purpose technical framework. They require human judgement informed by professional experience.

AI polish can be highly deceptive. As a veteran comms professional of 30 years, I have often found myself approving an AI output at first only to revisit it later and discover many weaknesses. Meaning that when we rely solely on end-stage checks, we mistake the detection of poor wording for the mitigation of communications risk. This is not an end-stage risk. The risk was actually introduced much earlier, during the design and training phases, long before the final review took place.

This forces us to ask a vital governance question: What are we allowing the machine to decide?

The Invisible Risk Of Autonomous AI Content Creation

The real risk is being lulled into a false sense of security by a system that functions exactly as designed and still produces communication that is culturally tone-deaf, strategically weak, or quietly erosive to the brand reputation. This is especially important in Malaysia, or even Asia, since language and cultural nuances often colour communications in this part of the world. Add the incredible cultural diversity in this region, and suddenly one enterprise level solution seems to be more of a risk than a benefit.

The benefit of highly autonomous agents in the corporate communications process can be compelling: prompt the agent and received a polished output in minutes instead of hours or even days with humans. Humans need only to do the final review of an output.

The generic AI-generated communication with the missing ability to navigate multiple language and cultural nuances is stepping into a reputation landmine of its own making. Because if an AI system is weighted to prioritise speed over nuance, or if its underlying design lacks robust skills and context, it is building on a flawed foundation. Such an AI system has technical accuracy but cannot determine whether a message is culturally relevant, strategically sound, or safe for the brand.

Culture, stakeholder context, strategy, and brand risk cannot be reduced to technical compliance. These qualities depend entirely on contextual human judgement. They require the nuanced understanding of practitioners who understand the ground realities of the markets they operate in.

"Good' Communications Has To be Defined Upstream by Human Input

For corporate communications directors, upstream AI-in-communications involvement is not a supplementary best practice, it needs to be a strict governance requirement.

Communicators must sit at the AI system design table to define what "good" looks like in a communications context. That means helping to define training data, prompt architecture, evaluation criteria, and guardrails - not from a technical standpoint, but from a communications standpoint. It means ensuring that the machine is not making judgements about communication quality that no one has explicitly authorised to ensure that the technology is able to serve communications objectives rather than invisibly undermining them.

Communicators must also have a well designed process to define the inputs going into the AI process to ensure communication outputs. Without robust human inputs into an AI system, including well-defined training data, prompt engineering, and guardrails, the system will often make a series of critical strategic choices autonomously, without the expertise that distinguishes competent communicators from generic AI sloppery.

Even if we correct AI's problematic output at the end stage to ensure strategically sound communications, that means we are humanly revisiting the entire content production process which negates whatever gains that AI is supposed to provide.

What This Means for Corporate Communications Heads

The question to ask is not whether your enterprise AI system works. The question is whether it has been taught what communications success actually looks like — and whether the comms professionals who understand that question have been involved in the answer. Without this clarity, companies risk investing in platforms that generate AI slop that produces less-than-meaningful communications outcomes.