Sub-shift
“RELATIONSHIP MEDIATION”
Help with the conversation becomes part of the relationship.
The apology has been rewritten three times. The message to an old friend is still sitting in drafts. The family chat has taken a turn that nobody quite knows how to reverse. There are plenty of conversations people want to have and struggle to begin.
AI offers help at precisely that point: trying out an opening, translating a thought, finding a less clumsy way to disagree. Over the next one to three years, we expect some consumers to make that assistance part of the preparation for everyday human exchanges. The benefit they seek is a conversation they can actually enter, with more confidence about what they mean.
“RELATIONSHIP MEDIATION” follows this expansion from correcting language to helping people navigate one another. For communication services, it opens a more ambitious role—and a more sensitive one—than supplying the next few words.
Expression carries expectations
People have long asked friends for advice about what to say. An AI assistant makes a version of that rehearsal available privately and at any hour. It can offer several ways into an exchange without requiring the sender to reveal the situation to another person first. Whether that privacy is real depends on the service, of course, and on whose messages have been supplied to it.
The important difference emerges when advice becomes the message itself. A polished apology can communicate a feeling the sender genuinely has, or one they have scarcely considered. The recipient may care very much which it is. Assistance that helps someone express uncertainty could be more valuable than a confident sentence they would never choose unaided.
Hohenstein and colleagues' 2023 smart-reply experiments captured part of this tension. Actual use made brief task conversations faster and more positive and improved interpersonal perceptions, while suspected AI use attracted more negative evaluations. That is evidence about a particular communication setting, not proof that assistance repairs lasting relationships. [50]
Leave something for the other person to answer
If you build messaging, social or community products, consider the moments when a customer is avoiding an exchange. A useful intervention might help them decide what they want to convey before offering language. It could preserve a disagreement, explain an ambiguous phrase or let them rehearse a difficult opening without sending anything.
The commercial opportunity lies in helping people have conversations they would otherwise postpone or abandon. A warmer tone alone is a poor measure of that success. Someone may need to set a boundary, decline an invitation or admit that they do not yet know what they think.
Keep private practice distinct from speaking on a person's behalf. There is a considerable difference between helping a customer prepare for a conversation and letting an agent conduct it as though it were them. The other participant has expectations, too. A service that respects both sides could earn a place in an intimate part of daily life without claiming to know what either person ought to feel.
THE SOURCE NOTESFollow the thinking 1 source · what they show and where their limits lie
Sources and further reading
These sources inform the article. A product announcement describes an offer, not how widely people use it. Arguments and imagined futures are kept separate from observed results.
Brief conversations show language and perception effects, not long-term relationship outcomes.
About this research edition
This article draws on selected sources and our interpretation of their implications for consumers. The wider collection contains 50 articles and original illustrations. We have not read everything by everyone on the reading list, calculated the probability of each future or established worldwide adoption. The collection is not exhaustive.
The illustrations are AI-generated, with a visual approach informed by TrendWatching’s 2024 and 2025 reports. They depict ideas rather than observed events. If an article’s central idea changes, its image needs another look. Small edits need not mean a new illustration; automatic replacement is not active.
Each innovation has a source, a review date and an explanation of its relevance and limits. Some are early signs of the change; others offer a useful comparison. None proves the forecast. Monthly monitoring and automatic selection of new examples are not running.