Relying too much on AI impedes growth of skills essential for effective communication

DAR ES SALAAM: IN a previous discussion published in the Daily News, concerns were raised about what Tanzania needs to do, to avoid “digital dementia”, particularly as artificial intelligence (AI) has become increasingly common across the private and public sectors, NGOs and even within families.

The discussion highlighted an ongoing debate fuelled by rapid developments in AI, which have also revived concerns about its potential risks.

Industry experts have warned about the possibility of increasingly advanced AI systems escaping human control and posing risks to humanity.

This raises an important question: Are responsible companies doing enough to prevent such possibilities? AI is rapidly changing how we learn, work, communicate and solve problems, particularly through increasingly sophisticated applications and programmes.

In Tanzania, as in many other countries, AI offers significant opportunities to transform education, improve productivity and prepare young people for an expanding digital economy.

However, there is also a significant risk that excessive reliance on AI could weaken independent thinking as technology increasingly takes over tasks that require human judgement. It could also undermine important communication skills.

A colleague once asked me to review a letter he had drafted using an AI writing tool.

It was perfect, grammatically correct, well organised and professional. Yet, something seemed amiss. He was writing to a long-term client about a delayed project payment.

The letter was polite, but the reminder felt as though it had been written by someone unfamiliar with the client or their relationship.

It contained all the right phrases but lacked the personal touches that would have made it feel authentic.

This is happening increasingly as AI writing tools become ubiquitous in professional and consultancy settings.

Many people, often without realising it, now rely on these tools to draft emails, prepare presentations, produce reports and improve their writing.

They are highly effective at transforming rough ideas into polished text, adjusting tone, correcting grammatical errors and organising arguments clearly. There is no doubt that such tools are useful for many everyday communication tasks.

However, their limitations also demonstrate that they cannot completely replicate human interaction, highlighting what makes human communication unique. AI is highly effective at basic language tasks.

It can produce clear, accurate and well-structured sentences. Its training on vast amounts of text enables it to handle grammar and language patterns at a level that may surpass many individual users.

It can imitate different styles, adjust levels of formality and generate fluent, professional writing. If communication were solely about encoding information into grammatically correct sentences, AI would come close to being perfect.

ALSO READ: Protecting critical thinking in Artificial Intelligence age

But communication is not simply about encoding information. It also involves managing relationships, understanding context and deciding what to say or leave unsaid, based on the person listening and the purpose of the interaction.

These elements require more than the ability to identify patterns in text. Consider what happens when you ask AI to generate an apology letter.

It will usually include the expected elements: Acknowledging the problem, expressing regret and suggesting some form of remedy, all in appropriate language.

But it will not necessarily know whether this is your first disappointment with the person or your fifth. It may not know whether your relationship can withstand a straightforward apology or whether a deeper issue needs to be addressed.

It may not know whether the person prefers direct communication or values a more detailed expression of remorse. Although these factors can determine how an apology is received, they extend beyond the words themselves.

This is not simply a problem that can be solved by providing better training data. It reflects a fundamental challenge in using AI for human communication: Meaningful communication depends heavily on context.

When we speak to another person, we draw on shared history, common experiences and an understanding of what matters to them. We decide what to state directly and what to communicate indirectly.

We observe how our message is received and adjust our approach accordingly. These are abilities that AI cannot independently reproduce in the same way a human communicator can.

The limitation becomes particularly clear in AIgenerated professional correspondence. For example, a manager might use AI to draft a concept note explaining a policy change. The document may provide a straightforward and well-structured explanation of the update and its rationale.

However, AI may not know which team members are likely to be concerned about particular details, recognise unspoken anxieties or understand how the proposed change relates to previous decisions that did not go well.

A human manager who understands the team’s dynamics would naturally take these considerations into account, not because they are following a communication template, but because they understand the relationships and history behind the communication. The same limitation applies to tone.

AI can modify the level of formality, making a message friendlier, more authoritative or more concise. But human tone involves much more than word choice.

It can signal the relationship between two people, the speaker’s attitude towards the subject and the degree of confidence or hesitation behind a statement.

When a colleague says, “I think we might want to reconsider this approach,” the hesitant wording may signal their level of confidence, their relationship with you and how comfortable they feel challenging your ideas.

AI can generate such hedging phrases, but it cannot always determine when hedging is appropriate or when it might weaken the message. That decision depends heavily on context.

This raises an intriguing question: What truly constitutes effective communication? We often think of good communication as the ability to express ideas clearly. That is certainly one of its essential elements, and it is an area in which AI performs well.

But human communication involves much more. It requires understanding what the listener already knows and does not know, what they care about, what might concern them and how they are likely to interpret what you say.

It involves deciding what to emphasise and what to downplay according to your objectives and relationship with the audience.

It also involves observing responses and adjusting your approach accordingly. These are practical aspects of communication.

The field of pragmatics examines how context affects meaning. The same words can be understood differently depending on the circumstances in which they are used. Much of human communication is also implied rather than explicitly stated.

From this perspective, meaning does not exist only within sentences. It develops through interaction between the speaker and listener and is influenced by their backgrounds, relationship and circumstances.

AI is highly capable at the semantic level: It can identify the general meanings of words and generate grammatically correct and logically structured sentences.

But effective communication also requires pragmatic understanding, recognising what particular words imply for a particular person in a particular situation, and judging whether a statement is appropriate given the relationship and history be tween the people involved.

Technically correct language can still sound dismissive, insensitive or out of touch.

I have seen this pattern in many professional settings, including workshops and conferences. For instance, a consultant might use AI to develop a comprehensive corporate strategic plan for a client. The report may be detailed and well-constructed, offering recommendations that appear logical.

Yet some recommendations may overlook political, economic or organisational realities known to the consultant but not fully captured in the information provided to the AI system.

Similarly, a team member might use AI to respond to a sensitive letter from a loan applicant.

The response may be polite and address the key issues, but overlook emotional nuances and leave the client feeling that their concerns were not genuinely understood. A leader might also use AI to draft remarks for a town-hall meeting.

The remarks could be straightforward and cover all the essential points, yet lack the personal warmth that signals genuine engagement with employees’ concerns. In each case, AI has performed its immediate function by generating clear, accurate and well-organised text.

But effective communication requires more than well-constructed sentences. It requires attention to relationships, context and the particular audience.

It requires decisions about what to say, how to say it and, sometimes, what not to say. This is not an argument against using AI for communication tasks.

AI can genuinely help with drafting, editing and refining prose. It can save time, improve clarity and help people organise their thoughts.

The difficulty arises when AI-generated text is treated as complete communication rather than as a starting point that requires human judgement before it reaches its intended audience.

AI’s limitations highlight how much of human communication occurs at the pragmatic level rather than purely at the semantic level.

Effective communication involves more than accurately encoding information. It requires deliberate choices about how information should be presented based on an understanding of the audience.

We signal our relationship with others and our attitude towards the subject.

We consider how our words may affect the other person and respond to their reactions by adjusting our approach. We decide what to express explicitly and what to imply. We pay attention to emotional subtext and relational dynamics.

These are not optional additions that simply make communication more pleasant. They are essential to whether communication achieves its purpose.

A letter that is perfectly clear but ignores the relationship between the sender and recipient can damage that relationship even while conveying the information accurately.

A presentation that is logically structured but fails to address the audience’s actual concerns may fail to achieve its purpose, regardless of how well its arguments are written.

Likewise, a policy announcement that is comprehensive but insensitive to employees’ concerns may create resistance rather than understanding.

AI cannot reliably perform all these tasks because they require more than processing language.

They involve understanding the people involved, their interaction history, the social or organisational environment and the objectives of the communication.

They also require interpreting how a message is being received and adapting accordingly.

In addition, they require judgement about what is significant in a particular situation, judgement rooted in human values, experience and priorities. This is important when considering how AI should be used in professional communication.

AI performs particularly well on tasks with limited or clearly defined contexts.

It can improve grammar and clarity and produce preliminary drafts that users can then refine using their knowledge of the specific situation and relationships involved.

It can also handle routine communication where relatively little contextual or relational judgement is required.

But it should not be treated as a substitute for human judgement, particularly when communication involves important relationships, sensitive issues or complex social contexts.

Perhaps the most important lesson AI reveals about communication is how much of it we have been doing unconsciously.

Writing a letter to a colleague involves more than sharing information. It draws on our understanding of that person, our relationship with them and the circumstances surrounding the communication.

These factors influence our decisions about tone, framing and what to include or omit. When speaking to a group, we interpret the room, recognise signs of engagement or confusion and adjust our delivery.

Handling difficult conversations requires us to manage multiple layers of meaning while balancing the content of what we are saying with the relationship between those involved.

These are advanced cognitive and social skills that we use constantly, often without consciously recognising them. AI’s limitations in performing such tasks reveal their complexity.

They remind us that effective communication is not only about producing clear sentences. It is also about context, relationships and the pragmatic factors that determine whether a message achieves its intended purpose.

For professionals seeking to be recognised as experts in written communication, the lesson is clear: AI can be a valuable aid for drafting, editing and refining communication, but human judgement remains essential in deciding what to say, how to say it and when to remain silent.

Those decisions depend on relationships, context, experience and understanding, not simply on the words themselves.