Luzmo research · 2026

The Everywhere Index

We asked software product teams where their customers get answers from product data today, where that is heading, and whether those answers can be trusted. The short version: customers now ask everywhere. In the dashboard, in a chat, in Slack and in their own AI assistant.

  • 62 software product teams
  • Responses from 15 September 2026 to 13 October 2026
  • Compared with our 2025 survey
Read the six findings Benchmark your team

74% of product teams say customers have asked to reach their data from their own AI assistant.

Only 31% of teams have shipped it

Highlights

Six findings, and how we read them

For each one: what we found, whether we saw it coming, and our take. Every chart links to the full data below.

74% of product teams say customers have already asked to reach their data from their own AI assistant. 31% have shipped it.

Expected the demand, not the scale

We expected demand, but not this much of it: in 47% of teams, several customers have asked, not just one or two.

Most roadmaps haven’t caught up. For a growing share of customers, their own assistant is where the next question gets asked, and the dashboard is no longer the default way in.

Agents will add to this: 45% expect them to make more than half of all queries on their data within two years. Given how much more often agents query than people do, we think that estimate is on the conservative side.

Our takeAccess is becoming as important as the interface.

Access from customers’ own AI

Via MCP, an API or a connector

Bar chart: access from customers’ own AI assistant, share of product teams.
AnswerShare of respondents
Customers have asked for it74%
Several customers have asked47%
Building it now or on the roadmap40%
Already shipped it31%

Select an answer to filter every other chart.

79% of product teams serve a dashboard today. Only 23% would keep it if they could serve answers in only one place.

Surprised us

We don’t read this as the end of the dashboard. It is still the surface most teams serve and expect customers to keep using. Its role is shifting: from the place people go for answers to the verified source that chats and agents draw on.

The split runs through the team: 33% of people who build analytics would keep the dashboard, against 17% of people who decide what gets built.

Our takeThe dashboard is becoming the foundation, not the front door.

If you could serve only one surface well for 12 months

Bar chart: the one surface product teams would serve well for the next 12 months.
AnswerShare of respondents
A chat or assistant inside your product34%
A dashboard inside your product23%
Their own AI assistant (ChatGPT, Claude, Copilot)19%
Their own BI tool, via API or data share8%
Inside the workflow, with no separate analytics screen8%
Pushed into Slack or Teams6%
Email or scheduled reports2%

Select an answer to filter every other chart.

33%of people who build analytics would keep the dashboard
17%of people who decide what gets built would

Only 13% of product teams are certain their dashboard, chat and AI agent give the same answer to the same question.

Surprised us

Every new surface is another route to the same number, and few teams test those routes against each other. 26% don’t know whether the answers match.

Among teams that shipped AI analytics, 66% have seen it give a real user a wrong or misleading answer. Only 5% are sure it never has.

Our read: AI shipped faster than the testing around it. Few teams evaluate AI answers the way they test other features, and the teams that know least about their AI’s mistakes are the most willing to let it answer unreviewed (64% vs 16%).

Our takeEvaluating AI answers belongs in the definition of done.

Same question in two places: do the answers match?

Bar chart: how sure product teams are that the same question gets the same answer in two places.
AnswerShare of respondents
Certain they match13%
Probably match47%
Do not know26%
Probably not13%
We know they do not2%

Select an answer to filter every other chart.

64%of teams not sure their AI was ever wrong would let it answer customers with no human review
16%of teams who know it has been wrong would

85% believe AI can surface insights they’re missing. But 44% of teams that shipped AI analytics don’t measure its use, or see fewer than 5% of users touch it.

Partly expected

Belief keeps rising: 85% think AI can surface insights they’re missing, up from 74% last year, and 66% have shipped something.

Usage hasn’t followed. Many AI analytics features seem to have been launched to be on the roadmap, without a clear owner for adoption. The main reported win is speed; very few teams can charge for it.

Our takeAdoption, not availability, is the metric that matters for AI analytics.

Share of monthly active users who used the AI analytics feature last month

Teams that shipped AI analytics

Bar chart: share of monthly active users who used the AI analytics feature last month.
AnswerShare of respondents
Under 5%20%
5-10%12%
11-25%15%
26-50%10%
Over 50%20%
We don’t measure it24%

Select an answer to filter every other chart.

66%say users now get answers faster
7%say they can charge more for it

63% of teams shipped their customer-facing analytics within about four months, up from 40% in 2025. Yet 20% say it’s never really finished.

Partly expected

We expected AI coding tools to speed up the first release, and they did: builds over six months fell from 18% to 8%.

What they don’t change is the work that follows: new requests, new clients, new surfaces. The “never finished” group didn’t shrink (19% → 20%), which suggests teams underestimate that part.

Our takeThe cost of building analytics sits in year two, not in version one.

Time to build customer-facing analytics

Grouped bar chart: time to build customer-facing analytics, 2025 compared with 2026.
Answer20252026
Within about a quarter40%63%
Four to six months23%8%
Longer than six months18%8%
Never really finished19%20%

Select an answer to filter every other chart.

47%are more confident building it themselves than a year ago
27%lean more towards buying a platform
2025 bands: under 3 months, 4–6 months. 2026 bands: under 4 months, 5–6 months.

48% say security, privacy or compliance is what holds back AI in their analytics. Accuracy and hallucinations come second at 40% .

Surprised us

Hallucinations get the headlines, but security, privacy and compliance top the list, and 32% say a compliance concern about a vendor changed a decision in the last year.

The concern is real, especially once customers’ own AI can reach product data. We suspect it also stands in for broader readiness: messy data comes third (31%), and the two tend to travel together.

Our takeReadiness, more than regulation, is what holds most teams back.

What is actually holding AI back in your analytics?

Up to three answers

Bar chart: what holds AI back in product teams’ analytics, top six answers.
AnswerShare of respondents
Data security, privacy or compliance48%
Accuracy and hallucination risk40%
Our data isn’t structured or clean enough31%
No clear user demand24%
Nothing significant24%
Lack of in-house expertise18%
32%say a security or compliance concern changed a vendor decision this year
15%built it themselves instead

One year later

2025 vs 2026: what moved and what didn’t

Both years in one chart per question, with what we expected and what we didn’t.

Dashboards help them make better decisions

Rated 4 or 5 out of 5

Dashboards help them make better decisions: 2025 compared with 2026.
Answer20252026
Dashboards help them make better decisions54%45%
Went down

We expected a year of AI features to lift this. It fell instead. We think dashboards have reached a ceiling, and the next gains in usefulness will come from other surfaces.

AI can surface insights they’re missing

Rated 4 or 5 out of 5

AI can surface insights they’re missing: 2025 compared with 2026.
Answer20252026
AI can surface insights they’re missing74%85%
Expected

Belief kept climbing, in line with a year of AI launches across the industry.

Shipped customer analytics within about a quarter

2025: under 3 months · 2026: under 4 months

Shipped customer analytics within about a quarter: 2025 compared with 2026.
Answer20252026
Shipped customer analytics within about a quarter40%63%
Expected

AI coding tools shortened the first build.

Say their analytics is never really finished

Say their analytics is never really finished: 2025 compared with 2026.
Answer20252026
Say their analytics is never really finished19%20%
Surprised us

Faster builds didn’t make analytics easier to keep up to date. The long tail of customer requests stayed.

Want predictive analytics

2025: what users ask for · 2026: what would most improve analytics (up to three)

Want predictive analytics: 2025 compared with 2026.
Answer20252026
Want predictive analytics38%42%
Expected

Predictive overtook real-time as the top wish, as AI made forecasting feel within reach.

Want real-time insights

Same question pair as above

Want real-time insights: 2025 compared with 2026.
Answer20252026
Want real-time insights51%35%
Surprised us

Last year’s top request dropped. Part of this is the narrower 2026 question; part, we think, is that “fast enough” data has become normal.

Regularly work around the dashboard

2025: export to spreadsheets or BI daily or occasionally · 2026: bypass dashboards always or regularly

Regularly work around the dashboard: 2025 compared with 2026.
Answer20252026
Regularly work around the dashboard72%61%
Directional

Fewer people work around dashboards, but the two questions differ, so read this as a direction, not a measurement.

Comparisons are directional: the two surveys reached different respondents and a few questions were worded differently. Where wording changed, the question is shown next to the chart.

Benchmark yourself

How ready is your product to answer everywhere?

Four questions, about a minute. Your answers stay in your browser.

Where do your users get answers from your product’s data today? Question 1 of 4 · Select all that apply · percentages show how many teams answered the same

Took the survey? You don’t need to fill this in. We’ll send your personal benchmark, based on your full answers, directly to you once the survey closes.

All the data

Every question, both years

Pick any question and split it by role, company size or another answer. Or ask in plain words with the Ask the data button at the bottom right.

About this research

Who took part

  • 62 software product teams completed the 2026 survey, between 15 September 2026 and 13 October 2026.
  • 55% product, 15% founder / ceo, 21% engineering & data, 10% other.
  • 129 people answered the 2025 survey, “Why do most dashboards get ignored?”.

How to read the numbers

  • Percentages use the people who answered each question.
  • Multi-select questions add up to more than 100%.
  • AI adoption and “what AI changed” are asked only of teams that shipped AI analytics.
  • “Rated 4–5” is the share choosing 4 or 5 on a 1–5 scale.

Comparisons and privacy

  • Year-on-year comparisons are directional: different respondents, and some questions were reworded.
  • Responses are anonymous. Roles and company sizes are grouped; no names, emails or companies appear here or in the data behind the charts.
  • Charts are built with Luzmo Flex on the anonymised survey dataset.

Not in the index yet?

Add your team while the survey is open. Six minutes, anonymous, and you get your personal benchmark first.

Take the survey (opens in new tab)
QR code that opens the survey