Tendances communication interne : ce qui change en 2026

Internal Communication in 2026: Three Shifts That Change What You Have to Prove

By Nicolas Saliba · 8 min read

In a Q1 2026 review, a Head of Internal Communication presented a reach figure that had appeared on the same slide for three years. This time someone from the people analytics team had the same population in a different system and got a different number. The rest of the meeting was about reconciling two spreadsheets rather than about whether the strategy landed.

That meeting is the trend. Not a new channel, not an AI feature. The thing that changed is where internal communication numbers now go: out of the monthly deck and into systems and conversations owned by other people. Numbers that only ever had to persuade get treated very differently once they have to agree with something else. Three shifts follow from that, and they change what you have to be able to prove.

Shift one: your numbers left the deck

For most of the last decade, internal communication measurement was a self-contained artefact. The team produced it, the team presented it, and nobody outside the function held a competing version. That is ending, for an ordinary reason: HR, people analytics and transformation offices are building dashboards that want communication activity as an input, and once your figure sits next to an engagement survey result and an HR headcount, it gets checked.

The requirement that emerges is reconcilability, which is a different property from accuracy. A number can be accurate and still be indefensible if you cannot explain why it differs from the number in the next column. Reach counted as page views will not agree with reach counted as distinct people. Neither is wrong. Both look wrong when they appear side by side without a stated definition.

Practically, this means writing down the definition of every headline figure, including the population it is calculated against and the deduplication rule, and being able to produce it on demand. Our companion piece on what executives started asking in 2025 covers the denominator question that opened this, and the 2026 benchmarks note explains why external comparison is only useful once your own definitions are stable.

Why Power BI is where this usually surfaces

The reconciliation problem tends to arrive through a dashboard build. Teams point Power BI at the Microsoft Graph reporting API, publish a dashboard alongside the HR one, and discover the difficulty at the worst moment. Be clear-eyed about it: the reporting API is not consistently reliable and produces days with missing data^[1]. Customers have observed KPI values in Power BI that do not reconcile with the native SharePoint reports for the same period. A dependable build needs a robust ingestion process, scheduled reconciliation against the native figures, and a named owner. Budget for the pipeline rather than the dashboard.

Shift two: content got cheap, so attention became the audited quantity

Generative tools removed the production constraint. A team of four can now publish what a team of ten published in 2023, and many have. The consequence is not that quality collapsed. It is that volume stopped being evidence of anything.

For twenty years, output was an acceptable proxy for effort, and effort was an acceptable proxy for value. That chain has broken at the first link. When your executive sponsor knows a draft takes twenty minutes, “we published 340 items” reads as a cost, not an achievement. Deloitte’s human capital research has repeatedly framed this period as one where organisations must decide what human judgement is for once the routine production is automated^[2], and internal communication is squarely inside that question.

The measure that replaces volume is attention: how many distinct people, in which parts of the organisation, engaged with the small number of messages that actually mattered. That is a harder number to produce and a much easier one to defend. The note on AI sentiment analysis covers where machine reading genuinely helps and where it produces confident nonsense.

Shift three: the audience definition is now contested

The third shift is the one with no settled answer yet, which is why it belongs in a forward-looking piece rather than a retrospective.

Who counts as reached is becoming genuinely ambiguous. Three populations sit at the edges. Deskless and frontline workers, many of whom have no corporate mailbox and reach the intranet through a shared terminal, where Gallup’s workplace research keeps finding the widest engagement gap^[3]. Long-term contractors who do the work but sit outside the tenant. And now a third: people whose assistant read the message for them.

That last one is new and is going to matter. When someone asks Copilot what happened this week and gets a summary drawn from intranet news, they have been informed without generating a page view^[4]. Every measurement framework in internal communication assumes the reader visits the content. Assistant-mediated reading breaks that assumption quietly, and the first visible symptom is a slow unexplained decline in page views against stable or rising comprehension. Our piece on Copilot and internal communications measurement works through what can and cannot be observed.

2026 shift	What it makes obsolete	What it requires instead
Numbers read outside the function	Metrics defined informally inside the team	Written definitions, stated denominators, reconciliation
Content production became cheap	Volume as evidence of contribution	Distinct attention on the messages that matter
Audience boundaries contested	Headcount as the universal denominator	An explicit, agreed addressable population
Assistant-mediated reading	Page views as a proxy for informed	Comprehension checks and cross-surface reach

What has not changed, and should not

A trends piece that claims everything is different is not worth reading, so here is the concession, properly made.

The craft is unchanged. A clear message from a leader who is trusted still outperforms a well-instrumented campaign from one who is not, and no amount of segmentation fixes a message that has nothing to say. Editorial judgement about what deserves the organisation’s attention is still the scarce skill, and it is the part generative tools are worst at, because it depends on knowing what happened in the last reorganisation and who is still angry about it.

Channel proliferation is also not a trend. It is a permanent condition, and it has been for a decade. The organisations that handle it well are not the ones with the newest platform. They are the ones that decided which surface carries which kind of message and then held that line, which is the argument our audit of the 2025 trend lists made against most of what was published that year.

What to put in the 2026 plan

Four things, in order of how much argument they save you later.

Write the definitions down. Every headline metric, its population, its deduplication rule, and its source. One page. It will be read by someone outside your team within the year.
Agree the addressable population with HR. Not headcount. Who you are accountable for reaching, including or excluding contractors and deskless staff, agreed and dated.
Pick the ten messages that matter. Measure those properly and stop reporting on the other 330. Gallagher’s State of the Sector research keeps identifying measurement as the capability internal communicators name first as missing^[5], and part of the gap is simply that teams try to measure everything.
Instrument for assistant-mediated reading now. Even if you can only observe it indirectly, knowing the trend line exists is better than explaining a decline you did not anticipate.

This is the layer Tryane occupies. It reads SharePoint, Viva Engage, Teams and your newsletter platform together, joins that activity to your organisational structure through Entra ID or an HR file, deduplicates so a distinct-people figure is actually distinct, and keeps history across product transitions so a multi-year trend survives the next migration. Tryane is SOC 2 Type 2 certified and GDPR compliant by design, hosts in the EU by default with US data residency on request, deploys in a couple of hours with single sign-on through Entra ID, and installs no agent. If you are formalising the function this year, the maturity model is the right companion read.

Frequently asked questions

Is AI the biggest internal communication trend of 2026?

AI is the biggest change to how content is produced and a modest change to how it is measured. The larger shift is who reads your numbers. A team that adopts every AI tool and cannot reconcile its reach figure with the HR dashboard is in a worse position than one that does the reverse.

How do we measure whether Copilot summaries reached people?

Not directly, and anyone claiming otherwise is overselling. What you can do is watch for the signature: page views declining while comprehension checks, survey recall and downstream action hold steady. Treat that pattern as a signal about channel behaviour rather than a failure, and say so before someone else interprets it for you.

Should we still report reach?

Yes, with a stated denominator and a deduplication rule. Reach without a population is the metric that gets contested the moment it is read outside your team, and the contest is usually lost on definition rather than on substance.

How many metrics should an internal communication team report?

Fewer than most report now. A small set that is defined, reconcilable and consistently produced beats a large set that is none of those. Most teams could remove half of what they present and improve the conversation, because the removed half is generally the part nobody can define under questioning.

Do these shifts apply to a 2,000-person organisation?

The first two apply at any size. The third bites hardest in organisations with substantial deskless populations or heavy contractor use, regardless of headcount. A 2,000-person manufacturer has a sharper audience-definition problem than a 20,000-person professional services firm.