Instead of building a dashboard tile by tile, you can describe what you want in plain English and Tali will build it for you, working out which data to use, creating the charts, and laying them out on the page. Once it's built, you don't need to start again to make changes. Just tell Tali what to adjust and it will refine the existing dashboard rather than starting from scratch.

Building with Tali uses the same access as building a dashboard normally, the 'Can Manage Dashboards' permission. If you can already create and edit dashboards in Talos, you can build one with Tali.

Create a new dashboard as you normally would, then use the Tali panel to describe what you're after. The more context you give about who the dashboard is for and what it needs to help them see, the better Tali's first attempt will be.

For example: "Create this dashboard for my Talent Acquisition Manager to review vacancy performance based on applications and time to hire, so we can see which roles are struggling to attract candidates."

Tali automatically works out which of your reporting models are relevant to your prompt, so you don't need to select one yourself. That said, knowing the names of the metrics you care about (applications, time to hire, source performance, and so on) helps Tali build something closer to what you actually want first time.

Here's a real example of exactly this in action, starting from an Applicant Raw Data report sitting on its own on a dashboard, one row per application, with no charts at all.

The prompt used was: "Turn this into a dashboard showing total applications by source, by team and by current milestone."

From that single sentence, Tali built a full dashboard: bar charts breaking applications down by source, by team and by milestone, a funnel chart showing candidates flowing from Incomplete through to Hired, a ranked chart of the top 5 candidate sources, a written Insight Tile summarising the week's applications, and it kept the original raw data table underneath so you can still drill into individual records.

Tali's own summary of the work is worth a look too. It explained what each chart shows, grouped them logically into rows, and finished by asking two sensible follow-up questions of its own: offering to filter the top sources chart to one team, and offering to add dashboard-level filters so the whole thing can be sliced by team.

The funnel chart and the Insight Tile it added are a good example of Tali noticing something worth surfacing on its own, in this case a team's applications dropping to zero:

That Insight Tile, titled "Applications Dropped to Zero This Week, HR Team Needs Immediate Investigation", is a real Insight Tile doing exactly what's described below: reading the live data each time the dashboard loads and writing a plain-English flag, rather than leaving you to spot the drop yourself in a table of raw rows.

You don't need to get everything right in your first prompt. Open the dashboard again at any point and enter a further prompt to adjust it, the same way you'd ask a colleague to tweak something they'd already built for you.

For example, once your first version exists: "We also have location data for our vacancies, can you use that to show performance by region as well?"

Tali can make a range of changes this way, including:

Here are a few more examples of the kind of prompts you can use, at different stages of building a dashboard.

Starting a dashboard from scratch:

Refining an existing dashboard:

Adjusting formatting and layout:

Alongside charts, Tali can add an Insight Tile to a dashboard. This is a tile that shows AI-generated written analysis instead of a chart. You give it instructions for what to look for, and it runs those instructions against your live data every time the dashboard loads, so the commentary stays current rather than being written once and left to go stale.

To add one, open the dashboard, select More, then Edit, and ask Tali for an insight or summary tile through the panel. Tali will check with you which chart, tile or tab you want it to focus its analysis on before generating it.

For example: "Add an insight tile that summarises how this week's applications compare to last week, and flags it if any team has had a drop of more than 20%."

The worked example earlier in this guide shows a real Insight Tile built from close to that exact prompt. Titled "Applications Dropped to Zero This Week, HR Team Needs Immediate Investigation", it read the live applications data, spotted that a team's applications had fallen from 5 to 0 week on week, and wrote that up in plain English rather than leaving it to be spotted by eye in a chart.

An Insight Tile can be as simple as a single summary sentence, or a more detailed, multi-step analysis, depending on how much detail you give it in your instructions. As with cross-model dashboards generally, Insight Tiles respect your existing security setup, so a viewer only ever sees commentary based on the data they're permitted to see.

Building with Tali is still developing, and a few things are worth knowing before you rely on it:

Do I need to know exactly which reporting model to use before I start?
​

No. Tali automatically identifies the most relevant data for your prompt, though being specific about the metrics and terms you care about helps it build something closer to what you want first time.