> For the complete documentation index, see [llms.txt](https://docs.startdeliver.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.startdeliver.io/jecta-agent-and-ai/ai-assessments.md).

# AI Assessments

Every customer assessed automatically. No manual scoring. A current read on every customer, split into the areas that matter, with a recommended next step.

**Where to find it:** Every customer profile → Assessment panel · Settings → AI → Customer Intelligence\
**Who can access it:** Team members with access (set under *Who can see it*). Tuning: Admins

<figure><img src="https://1065233822-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fttka6LKL3bc1LsaDzuI5%2Fuploads%2FE4qDcloqs27WeT6tiADA%2FSka%CC%88rmavbild%202026-11-13%20kl.%2019.21.26.png?alt=media&amp;token=3f7548d3-a482-44f3-b4e7-0651ea89e921" alt=""><figcaption></figcaption></figure>

***

**What it is**

AI Assessments is the intelligence engine at the core of Startdeliver. It reads all available context for every customer (product usage, feedback, support activity, stakeholders, revenue data and more) and produces a plain-language assessment of where the customer stands and what your team should do next.

The assessment runs continuously and automatically. Nobody needs to trigger it or fill in a form. When new data comes in, such as a survey response, a delayed ticket, a drop in usage or a week of silence, the assessment updates.

***

**What the assessment produces**

On every customer profile, the Assessment panel shows:

**Status**: a plain-language verdict on the customer's current state, for example *Steady*.

**Summary**: one line on what is driving the status. For example: "Stable adoption in one team, no growth and no new goals."

**Trend**: how the assessment has moved over the past weeks and months.

**Take Action**: a recommended next step tied to what the AI found for this customer.

**The areas**: a short verdict per area (see below).

**See full analysis**: every signal the AI read, the data behind it, and the reasoning.

***

**The areas Jecta looks at**

The assessment is built from components. Each component reads one part of the customer relationship and gives a short verdict. On the customer profile they appear as **Jecta looked at 5 areas**:

| Area                 | What it looks at                                                                                                                                                          |
| -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Interactions**     | Meetings, emails, tickets and other touchpoints. Is the relationship active, recovering or going quiet?                                                                   |
| **Product usage**    | Adoption across products and users. Is usage broad and deepening, or narrowing?                                                                                           |
| **Stakeholders**     | Who you have at each customer across three roles (decision makers, super users and users), the relationship with each, and whether that is enough to defend the customer. |
| **External context** | What is happening around the customer, such as acquisitions, growth or cost pressure.                                                                                     |
| **Expansion**        | Whether there is a real foundation to grow the customer, and what is in the way if not.                                                                                   |

Example from one customer: *Interactions: Cordial, narrow · Product usage: Solid, narrow · Stakeholders: One thread · External context: Funded growth · Expansion: Qualified opportunity.*

**Open an area** to see its headline, its signal (for example *Good signs* or *Concerning*), what is working, what to watch and the topics it is based on. From there you can:

* **Refresh** the area if you know something has changed.
* **Want to adjust assessment?** opens Tune Studio for that component (see [Tuning & AI instructions](https://docs.startdeliver.io/jecta-agent-and-ai/ai-instructions.md)).
* **Ask Jecta** a question about it, or **Open in chat** to continue in the full chat. See [Jecta: AI Agent](https://docs.startdeliver.io/jecta-agent-and-ai/jecta-ai-agent).

If an area says *No assessment yet*, the customer is either not in the selection for assessments (see *Customer scope*) or the assessment is still pending.

<figure><img src="https://1065233822-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fttka6LKL3bc1LsaDzuI5%2Fuploads%2Fe9Ml7mKBoQJ62qOLLXkK%2Farea-popover.png?alt=media&amp;token=7b65e51d-2930-4450-bb0e-0bedf62594f2" alt=""><figcaption></figcaption></figure>

The AI does not average these areas into one score. It reads them together. A customer with solid usage in one team, a single contact and funded growth is a specific situation, and the assessment says what it means.

***

**Flags the AI raises**

Beyond the general assessment, the AI flags specific patterns:

**Churn risk**: early warning signs such as declining usage, poor feedback, unresolved support issues or communication that has gone quiet.

**Expansion opportunity**: strong adoption, positive sentiment and growing usage. The customer may be ready for an upgrade, another product or a wider rollout.

**Reference and advocate**: consistently strong feedback across several contacts. Worth a conversation about a case study or reference call.

**Silent customer**: no engagement across any channel. Not necessarily churning, but worth attention before the silence becomes a decision.

***

**In the customer list**

The assessment and each area can be added as columns in the customer list. Each cell shows the short verdict with a coloured dot, so you can scan every customer at once and see where to start.

<figure><img src="https://1065233822-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fttka6LKL3bc1LsaDzuI5%2Fuploads%2FQa9bLlbdHQfhy3FOy8Gh%2Fcustomer-list.png?alt=media&amp;token=eaf46add-4ad9-4982-915a-e57817a870ba" alt=""><figcaption></figcaption></figure>

***

**Tuning the assessment**

The assessment works out of the box. To make it fit your business, there are two layers:

**AI instructions (global).** Written guidance that applies to every assessment and everything Jecta does in the workspace. Use it for segment exceptions, commercial thresholds or your own terminology.

**Tune Studio (per component).** Admins can change how each component reasons, test the change on preview customers and apply it. See [Tuning & AI instructions](https://docs.startdeliver.io/jecta-agent-and-ai/ai-instructions.md).

***

**Other settings**

Under Settings → AI you also control:

**Who can see it**: *Everyone* on your team, or *Limited* to the team members you pick. Others don't see Customer Intelligence on customer pages, in columns or in filters.

**Customer scope**: which customers get assessed. Useful if you roll out segment by segment.

**Interaction settings**: which interactions are included in the context, plus custom interaction types for your business.

**Health signals** (e.g. product usage goals per product) also feed the assessment. See [How to set up health metrics & assessment](https://docs.startdeliver.io/health-and-metrics/how-to-set-up-health-metrics-and-assessment).

***

**Credits**

Assessments and the other AI features in Startdeliver run on credits. Go to **Settings → AI → Usage & Credits** to see:

* **Credits used this month** against what your plan includes.
* **Estimated usage**: what your current setup is likely to use per month. Check it before you run a tune for every customer or raise the *Effort* level.
* **AI Plan**: every Startdeliver plan includes AI credits (*Essentials*). For more, upgrade to *Accelerate* or *Frontier*, or contact us for a custom plan.
* **Top up credits**: one-off credit packs that carry over until used. Monthly plan credits reset on the 1st.

Credit use depends on how many customers are assessed and the *Effort* level per component: Light uses the fewest credits, Extensive the most. See [Tuning & AI instructions](https://docs.startdeliver.io/jecta-agent-and-ai/ai-instructions.md).

<figure><img src="https://1065233822-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fttka6LKL3bc1LsaDzuI5%2Fuploads%2FOhP11XfiQI2Req7FAuHW%2Fusage-credits.png?alt=media&amp;token=4996c306-ce5e-43ac-bda2-99450c0fb6f4" alt=""><figcaption></figcaption></figure>

***

**Where to go next**

→ [Tuning & AI instructions](https://docs.startdeliver.io/jecta-agent-and-ai/ai-instructions.md): global instructions for your business and how each component reasons\
→ [Jecta: AI Agent](https://docs.startdeliver.io/jecta-agent-and-ai/jecta-ai-agent): ask Jecta about any customer and get drafts to approve\
→ [MCP](https://docs.startdeliver.io/jecta-agent-and-ai/mcp-model-context-protocol): use your customer data in Claude and ChatGPT
