Business context for AI

Give AI the context
of your business.

Make business data, definitions and working practices available together to your AI assistant. With ThinkBox, you manage what is available, how data connects and who can use it.

ThinkBox provides the data and context.
Your connected AI assistant creates the answer.

A product by DevBox

A question for your AI assistantWith configured sources and access
Fictional example
You

Can I deliver everything I promised Delta on Friday? Check the customer agreements, order schedule and my email.

AI assistant

Delta needs 20 units for the opening. CRM

The schedule has 20 units on Friday and 60 on Tuesday. ERP

Your email promises all 80 on Friday. Outlook

The connection makes the difference

The schedule covers the opening, but not your full promise. Agree on a split delivery: 20 on Friday and 60 on Tuesday.

Business context via ThinkBox3 sources → 1 answer
Make business knowledge reusableRecord relationships between dataDecide what is available

The context behind the answer

Your business knows more
than a field can tell.

A customer is called a debtor elsewhere. Status A means ‘active’. And ‘resolved’ has exceptions. Record that knowledge centrally so your AI assistant can consult it for future questions.

01 / Data

Know what is available.

Manage sources and inspect samples of imported data. Make it clear which information you publish for use by AI.

Source catalogueData samples
02 / Meaning

Explain what you mean.

Give fields recognisable names. Describe terms, codes and business practices so the explanation is available alongside the data.

DefinitionsBusiness practices
03 / Relationships

Connect the whole story.

Record how customers, service requests and other data relate to each other. Explore suggested relationships and review which connections are valid.

RelationshipsReviewed suggestions
04 / Access

Decide what can be used.

Configure publication, roles and personal access. Make deliberate choices about shared and user-specific data.

PublicationRoles and permissions

Business knowledge that lasts

Explain it once.
Make it reusable.

Manage working practices with a topic, applicable sources and validity dates. Prepare a change as a draft while the published explanation remains available.

Example of a business practicePublished
“Use the invoice date for monthly revenue. State whether cancelled invoices have been included.”

Available as guidance for the assistant. Your organisation owns the content and checks how AI applies it.

DraftPublishUpdate

AI in practice

From scattered information
to useful insight.

A customer meeting, a delivery promise or an open quotation. With configured sources and recorded relationships, your connected AI assistant can combine information from several systems.

01 / Prepare for a customer meeting

The opportunity is in the connection.

Three pieces of information about the same customer:

CRM Customer Noordlicht
An open opportunity concerns expansion to a second location.
ERP Order history
From 100 to 130 units per quarter: a 30% increase in volume.
Outlook Email & calendar
A customer meeting tomorrow. The customer has emailed to ask about delivery to two addresses.

ThinkBox context: the customer record, orders and email addresses belong to the same organisation.

Measure in your pilotPreparation time, missed talking points and incorrect customer matches.

AI chatIn a connected AI application
Fictional example
You

I am meeting Noordlicht tomorrow. What opportunity do you see when you combine our CRM data, orders and my email?

AI assistant

The expansion opportunity in the CRM is becoming more concrete: order volume has grown by 30% and the customer has emailed to ask about delivery to two addresses.

Combined insight

Discuss a proposal for both locations, including volumes and delivery arrangements, at tomorrow’s meeting. The order growth and the email give you a concrete reason to follow up the CRM opportunity.

CRMERPOutlook
Business context via ThinkBox

02 / Check a delivery promise

Does the schedule match the promise?

The order does not tell you what you promised:

CRM Delta customer agreement
At least 20 units are needed for the opening on Friday.
ERP Order for 80 units
Delivery schedule: 20 units on Friday, the remaining 60 on Tuesday.
Outlook Sent email
“We will deliver all 80 units by Friday.”

ThinkBox context: the customer agreement, order and delivery promise concern the same delivery.

Measure in your pilotIdentified discrepancies and time spent checking the delivery promise.

AI chatIn a connected AI application
Fictional example
You

Can we honour what I promised Delta? Compare the customer agreements, order schedule and my email.

AI assistant

Not in full. Your email promises all 80 units on Friday, while the ERP schedules only 20 for that day. Those 20 do meet the minimum needed for the opening, as recorded in the CRM.

Combined insight

The schedule covers the opening, but differs from your promise. Discuss 20 units on Friday and 60 on Tuesday with Delta before the customer expects all 80.

CRMERPOutlook
Business context via ThinkBox

03 / Follow up a sales opportunity

A proposal that fits better.

Extra context sheds light on a budget objection:

CRM Rivierzicht quotation
€18,000 for two teams. Status: awaiting budget.
ERP Quotation items
A separately orderable part of €9,000 for each team.
Outlook Customer email
“The first team starts now; the second does not start until next quarter.”

ThinkBox context: the quotation, orderable items and correspondence belong to the same sales opportunity.

Measure in your pilotUsefulness of the preparation and corrections needed by the account manager.

AI chatIn a connected AI application
Fictional example
You

How can I make my follow-up on Rivierzicht’s open quotation more relevant? Combine the CRM opportunity, ERP quotation and latest email.

AI assistant

The CRM records a budget objection at €18,000. The email says one team starts later, and the ERP shows that the two parts can be ordered separately.

Combined insight

Suggest ordering €9,000 now and €9,000 next quarter. This matches the start dates of the two teams. Ask whether phasing the order would help address the budget objection.

CRMERPOutlook
Business context via ThinkBox

Fictional illustrations of a conversation in a connected AI application. Required source connections, relationships, user-specific email data and the AI connection are configured and checked for each use case. The assistant creates the answer; you assess the result and carry out any follow-up actions.

For the questions other teams ask, too.

Start with a recurring task. Agree in advance what makes an answer useful.

Service

The background at your fingertips.

“Which previous issues and service agreements are relevant to this customer?”
What you need
Accessible service records, customer relationships and current service agreements.
Measure in your pilot
Search time, missing context and incorrect matches.

Management

A shared meaning behind the number.

“How many requests have been resolved according to our agreed definition?”
What you need
Complete, accessible data for the period and definitions of statuses and exceptions.
Measure in your pilot
Compare the assistant’s calculation with a verified reference result.

Onboarding

Knowledge you can pass on.

“What does ‘waiting for customer’ mean, and which working practices apply here?”
What you need
Published definitions and practices, accessible examples and a knowledge owner.
Measure in your pilot
Accuracy, remaining questions and corrections by experienced colleagues.

How it works

From your business systems
to an answer with context.

Each component has its own role. Together they connect available information to its use by AI.

  1. 01

    Business systems

    Where you work and record information.

    CRMERPOutlook
  2. 02

    SyncBox

    Retrieves the agreed data and delivers it to ThinkBox.

    About SyncBox
  3. 03

    ThinkBox

    Stores data and manages meaning, relationships and access.

    Where context comes together
  4. 04

    AI assistant

    Retrieves permitted information, analyses it and formulates an answer.

    Through a suitable connection

We configure source connections, refresh schedules and the AI connection for your situation. Standard AI access is read-only. An authorised administrator can temporarily enable help with configuration; this does not change customer records or source systems.

See how ThinkBox records the context

Behind the scenes

This makes combining data possible.

Record how a CRM customer appears in orders and emails. Add explanations of codes and business practices, and decide which data is available.

The connected AI assistant uses that context to bring together information relevant to a question.

From data to meaningFictional example
CRM / CUSTOMERS
customer_id 1042status A
ThinkBoxContext recorded
Meaning
Status A means: active customer.
Relationship
Orders and emails belong to the same customer.
Business practice
Compare delivery promises with the order schedule.
Access
Only available according to the configured permissions.

Useful context for your AI assistantThe assistant combines information and formulates the answer.

Set up with your AI assistant

From explanation to setup.
With help from AI.

Let your assistant help make sources understandable. It can define readable field names and code meanings, record synonyms, and create and activate relationships.

An authorised administrator temporarily enables these capabilities for their own connection. You can then have multiple definitions and relationships recorded from a single conversation.

Discuss the setup of your sources

You open the admin window.

From 5 minutes to 4 hours · stop earlier at any time

  1. Enable temporary administration

    Choose your own sign-in or one personal key. Existing permissions and the scope of accessible data still apply.

  2. Start a new conversation

    Have the assistant examine source information and data samples. It can develop suggestions and save changes directly in ThinkBox.

  3. Check the content

    ThinkBox validates changes; you remain responsible for their meaning. There is no separate approval step. Once the window ends, new admin actions are refused.

The admin window does not renew automatically. These functions cannot change customer records, technical import keys or access settings.

Deliberate choices

Your data.
Your choices about its use.

Good use of AI starts with clear agreements. About the information you make available, who can use it and who maintains the setup.

More about data and access

Publish with a purpose

Prepare sources and practices as drafts. Decide what to activate and which guidance is available. Deactivation stops publication but does not delete stored data.

Configure access by role

Choose shared or user-specific data and configure ThinkBox’s own permissions. Users assigned to the same data group can see the same data. Administrative roles may have broader access.

Keep the content useful

Assign owners for meanings, relationships and business practices. Review answers and adjust the context where needed.

Revoke access when needed

Deactivate users or revoke personal AI keys. Updated permissions apply to subsequent requests; information already received in an AI chat is not recalled.

A focused first step

Start with one question.
Make the value measurable.

Choose a recurring task, a subject owner and a small group of users. A pilot shows what works and what still needs attention.

  1. 01

    Choose what should improve

    Record the question and the current way of working. Agree in advance what counts as a good result and which decisions people will continue to make.

  2. 02

    Set up the context

    Check source data and refresh schedules. Define terms, relationships and access, then connect the chosen AI assistant.

  3. 03

    Compare and improve

    Measure time including review, accuracy and missing information. Weigh the results against setup, maintenance and AI usage.

Expand when both the results and ongoing management are workable.

Discuss your first use case

Good to know

Frequently asked
questions.

Can’t find your question?
We are happy to help

Is ThinkBox a chatbot?

ThinkBox manages the data and context that a connected AI assistant can use. The assistant interprets your question and formulates the answer. ThinkBox does not include its own chat interface in this product setup.

Which systems and AI assistants can I connect?

Data is supplied through SyncBox. We determine suitable source connections and an AI connection for each use case. A source or assistant is ready for use once its connection, access and data exchange have been configured and checked.

How up to date is the data?

That depends on the source and the data delivery configured through SyncBox. ‘Last received’ shows when ThinkBox received a delivery; it does not prove that the entire source has been fully updated. We agree what freshness and completeness your use case requires.

Who can see which data?

ThinkBox uses its own roles, permissions and publication settings. Data can be configured as shared or user-specific. Administrative roles may have broader access. Permissions from a source system are not automatically copied one-to-one.

Where is my data processed?

ThinkBox stores the supplied data for use by the connected AI assistant. Hosting, data location, retention and deletion are part of the setup agreements. Processing by the chosen AI provider also needs to be considered.

Will AI always give the right answer?

Quality depends on the source data, the context recorded and how the AI assistant uses it. ThinkBox supports better-grounded answers. A person still reviews results where needed. Recorded business practices guide the assistant and are not automatically enforced as rules.

Can AI help configure ThinkBox?

Yes. An authorised administrator can enable temporary administration for their own sign-in or one of their AI keys. The assistant can change source information, field meanings, alternative identifiers and relationships, including activating relationships. The sources involved must already be active. ThinkBox validates and saves changes without a separate approval workflow.

Choose 5, 15 or 30 minutes, or 1, 2 or 4 hours. Then start a new chat and reload the connection if needed. The assistant cannot enable or extend this period itself. Read how temporary administration works.

Can the assistant change orders or send emails?

Standard AI access is read-only. A question through ThinkBox does not change a customer record, ticket or order, or send an email. Temporary AI administration can change configuration and meaning inside ThinkBox; it does not broaden access to customer data.

Are relationships discovered automatically?

After you select two sources, ThinkBox can suggest relationships based on a limited portion of the data. You can see matches and possible duplicates. An administrator reviews the connection: a suggestion does not prove that all records belong together and does not repair incorrect source values.

Which language does the admin interface use?

The current admin interface uses English. Names, descriptions and business practices can be entered in your organisation’s language. The chosen AI application determines which languages it supports for questions and answers.

How do we get started, and what does it cost?

Start with one recurring question, a clear owner and a result you can verify. Together, we identify the necessary sources, configuration, AI connection and how to evaluate the pilot. We discuss costs and service based on that scope.

Let’s make it concrete

What could your team
spend less time figuring out?

Tell us what information you are looking for and where it is today. Together, we can explore a suitable first use case for ThinkBox.

info@devbox.nl+31 6 58888706You will speak with the DevBox team.

Tell us which question you want AI to answer and which systems you use.