SaleAI Customer Settings is the place to define reusable business context for Copilot tasks: who your company is, what it sells, which markets and customer types matter, and how outreach should sound. Good settings reduce repeated explanation and make research or drafting prompts more consistent. Poor settings can repeat a vague positioning statement, an outdated market assumption, or an incorrect product classification across many conversations.
A read-only inspection of the logged-in SaleAI backend on August 12-14, 2026 showed Customer Settings alongside New Chat, Read Memory, Allow Learning, attachments, a gpt-5.4 model selector, and channel shortcuts. The detailed dialog recorded company and website, product and HS code, product description, preferred markets, customer types, preferred channels or products, and default email language and tone. This guide explains how to configure those observed fields without assuming unverified automatic actions.
1. Decide what the saved context must improve
An AI sales assistant setup should solve repeatable problems, not collect every fact the company owns. Before entering information, choose the decisions the context should improve:
- Which products should a research task prioritize?
- Which countries and buyer categories belong in the target definition?
- Which terms should the assistant use or avoid?
- Which evidence should be requested before a company is treated as relevant?
- Which language and tone should a draft use by default?
The settings should help SaleAI interpret a request such as “find distributors for our industrial sensors in Germany.” They should not make the final qualification, compliance, or outreach decision on behalf of the team.
2. What the inspected SaleAI controls actually do
The verified SaleAI backend workflow shows three separate context controls in the Copilot area.
| Control | Observed role | Appropriate use | Do not assume |
|---|---|---|---|
| Customer Settings | Maintain reusable company, product, buyer, market, and email rules | Store reviewed business defaults | Every field is automatically applied to every module or action |
| Read Memory | Bring saved context into the current conversation | Reuse relevant approved context | That all saved context is current or appropriate for this task |
| Allow Learning | Let the conversation contribute to reusable memory | Preserve a genuinely reusable correction or preference | That every conversation should become long-term context |
Read Memory and Allow Learning are independent decisions. A user may want the current task to read approved context without allowing its temporary assumptions to become reusable memory. Conversely, a team may decide that a reviewed correction should be learned only after the task is complete.
3. Separate stable context, task inputs, and external evidence
The most important rule for SaleAI Customer Settings is to separate information by lifespan.
| Context layer | Examples | Where it belongs | Review frequency |
|---|---|---|---|
| Stable business context | Company identity, official website, core product family, service model | Customer Settings | Quarterly or after a business change |
| Controlled preferences | Priority markets, customer types, preferred channel, email language, tone | Customer Settings with an owner and review date | Monthly or campaign-cycle review |
| Task-specific input | One country, one product variant, a date range, a temporary exclusion, a campaign offer | Current prompt or module task | Every task |
| External evidence | Customs record, company page, employee profile, map listing, current regulation | Research result with source and date | Before each consequential use |
Do not save a one-week promotion as a permanent company description. Do not save one buyer's behavior as a universal customer preference. Do not treat an external search result as a stable fact until it has been verified and deliberately approved.
This three-layer model makes sales AI context easier to maintain and prevents prompt history from becoming an uncontrolled database. Use the same separation in B2B lead source tracking: saved defaults describe the business, while dated records preserve where a specific claim came from.
4. Configure company identity and the official website
Begin the customer profile configuration with the company name and official website. Use the name the team expects prospects to recognize, then note a legal or manufacturing entity only when it changes how research should interpret the business.
A useful company statement answers:
- What does the company make or provide?
- Does it manufacture, trade, distribute, design, or combine several roles?
- Which business identity should appear in sales communication?
- Which website is the authoritative source for product and company claims?
- Are there brands, subsidiaries, or regions that require separate context?
Avoid promotional claims such as “the world's leading supplier” unless there is current, reviewable evidence and the phrase is approved for public use. The saved company description should help an AI distinguish the business, not imitate a homepage slogan.
5. Enter products as a controlled product map
Product information should be specific enough to guide research but not so broad that unrelated companies match. Treat these product and buyer settings as a controlled map for sales assistant onboarding, not as a replacement for the product catalog.
Use a product map with four parts:
| Product field | Example structure | Why it matters |
|---|---|---|
| Product family | Industrial temperature sensors | Establishes the main category |
| Applications | Food processing, HVAC, machinery, energy equipment | Connects products to buyer situations |
| Important variants | Probe type, range, output, housing, certification option | Separates real product differences |
| Exclusions | Consumer thermometers, medical devices, unrelated IoT products | Reduces false matches |
If the company sells many product families, start with the one used in the next pilot. A compact, reviewed description is more useful than a catalog pasted into SaleAI Customer Settings without hierarchy.
6. Treat the HS code as a research aid, not a universal product identity
The detailed Customer Settings inspection showed a product and HS code field. An HS code can support customs-oriented research, but it must be handled carefully.
The World Customs Organization's Harmonized System overview explains the international nomenclature used to classify traded goods. National tariff schedules can extend classifications beyond the international six-digit level, and the appropriate classification can depend on product characteristics and jurisdiction.
Use the following entry pattern:
- Candidate HS code and description.
- Products or variants it covers.
- Markets where the classification has been checked.
- Person or source responsible for the classification.
- Last review date.
- Known alternative codes or exclusions.
Do not present an AI-suggested HS code as legal classification advice. Confirm material classifications with qualified customs or trade professionals where needed.
7. Write a product description that supports buyer questions
A useful product description helps the assistant answer high-intent questions such as:
- Which buyer types use this product?
- Which application or problem makes the product relevant?
- Which specifications affect supplier selection?
- Which documents, certificates, samples, or customization questions may arise?
- What should not be claimed without confirmation?
Structure the description as positioning, applications, decision factors, proof, and exclusions. For example:
We manufacture industrial temperature sensors for equipment and process applications. Research should prioritize distributors, OEMs, system integrators, and manufacturers that specify sensing components. Relevant decision factors include temperature range, output, housing, ingress protection, calibration, certification, samples, customization, MOQ, and lead time. Do not claim certification, accuracy, capacity, or delivery terms unless current evidence is provided in the task.
This is stronger AI sales prompt context than a list of adjectives because it tells the assistant how buyers evaluate the product and where evidence is required.
8. Narrow preferred markets into operating priorities
“Europe” or “global” is rarely enough. A preferred market entry should capture the practical boundaries that affect research and communication:
- Country or a tightly defined region.
- Languages used for research and outreach.
- Whether the team can serve the market commercially and operationally.
- Product or certification constraints.
- Priority, secondary, or watch-list status.
- Review date and owner.
Use no more markets than the sales team can genuinely support in the next operating cycle. Save stable priorities in SaleAI Customer Settings, then choose a single country or segment in the current task. This keeps a broad corporate ambition from diluting a precise search.
9. Define customer types by commercial role
Customer type should describe how the account participates in the market, not merely its industry label.
| Customer type | What to clarify | Typical evidence needed |
|---|---|---|
| Distributor | Territory, product range, channel coverage, technical capability | Website, represented brands, locations, relevant trade or company evidence |
| OEM or manufacturer | Product application, component use, technical requirements | Product pages, facility context, engineering or procurement roles |
| Retailer or marketplace seller | Assortment, price position, private-label need, geography | Storefront, catalog, brand structure, channel activity |
| Contractor or integrator | Project type, specification role, installation capability | Projects, services, certifications, local coverage |
| Importer or wholesaler | Import role, category relevance, downstream network | Customs evidence, company identity, product and market fit |
One company may play several roles. Use a primary type for the current sales motion and keep alternative roles as hypotheses until verified. An export sales assistant should not assume that every importer is a distributor or that every manufacturer is an end user. Apply the account evidence rules in the B2B lead qualification workflow before changing a CRM stage or choosing outreach.
10. Configure preferred channels, email language, and tone
The inspected dialog included preferred channels or products plus default email language and tone. These are drafting defaults, not permission to contact everyone found through that channel.
For preferred channels, state the purpose:
- LinkedIn for professional and employee context.
- Google Search for websites and public business evidence.
- Google Maps for local presence and location-based discovery.
- Customs Data for transactions, counterparties, and trade trends.
- Facebook or Instagram for public brand and market activity.
- Email Marketing for reviewed activation after sender and recipient checks.
For language, specify the default and fallback. For email tone, use operational instructions such as “concise, specific, respectful, no unsupported familiarity, no exaggerated claims, and one clear next step.” That is more reliable than a single word such as “professional.”
Keep product terminology, units, spelling variant, banned phrases, and signature rules in a short style note if the current fields do not capture them directly. When a task is activated, connect these drafting defaults to the measurement discipline in the B2B email campaign metrics guide.
11. Choose Read Memory and Allow Learning deliberately
Use this decision table before sending a prompt. These AI memory controls are a central part of responsible B2B sales AI configuration, because reading approved context and saving new context create different risks:
| Task condition | Read Memory | Allow Learning | Reason |
|---|---|---|---|
| Standard research using approved company and product context | On | Off initially | Reuse stable facts without saving exploratory conclusions |
| Sensitive, one-off, or experimental task | Off or limited by team policy | Off | Reduce irrelevant or inappropriate reuse |
| Reviewed correction to a stable product or market preference | On | On only after approval | Preserve a verified improvement |
| Buyer-specific analysis with temporary assumptions | On if relevant | Off | Prevent one account's context from becoming a company default |
| Context audit or troubleshooting | Compare on and off | Off | Identify whether saved context is affecting the result |
The NIST AI Risk Management Framework, commonly called the NIST AI RMF, is a voluntary framework for incorporating trustworthiness considerations into the design, use, and evaluation of AI systems. Applied here, its governance logic supports named ownership, documented limitations, reviewable changes, and ongoing evaluation of saved context.
The inspected backend did not expose a full memory audit log, retention schedule, deletion workflow, permissions matrix, or field-level memory scope. Those capabilities should be confirmed before a team writes them into policy.
12. Create a context approval and update routine
Assign one business owner and one operational reviewer for SaleAI Customer Settings. The business owner confirms product, market, customer, and claim accuracy. The operational reviewer checks whether the text is clear enough to guide tasks without creating unintended matches.
Use four statuses:
- Draft: proposed context, not ready for routine use.
- Approved: reviewed and allowed as a default.
- Review due: still usable for low-risk tasks but needs confirmation.
- Retired: no longer valid and should not influence new work.
Review immediately after a product launch, certification change, market withdrawal, website or brand change, distributor-policy change, or correction to an HS classification. Otherwise, use a quarterly review for stable company and product context and a monthly review for market, customer, channel, language, and tone preferences.
The broader NIST Privacy Framework provides a voluntary structure for identifying and managing privacy risk. It does not define SaleAI's product behavior, but it reinforces the need to decide what information is appropriate to retain, who is responsible, and how the organization handles changes.
13. Run a 7-day configuration pilot
| Day | Task | Evidence to retain |
|---|---|---|
| 1 | Enter company identity, website, and one product family | Approved source and owner |
| 2 | Add the candidate HS code, applications, variants, and exclusions | Classification note and review status |
| 3 | Select one market and two customer types | Commercial rationale and exclusions |
| 4 | Set channel purpose, email language, and detailed tone rules | Style and channel checklist |
| 5 | Run the same research prompt with Read Memory on and off | Differences in relevance and unsupported assumptions |
| 6 | Review whether any task insight deserves Allow Learning | Approved reusable correction only |
| 7 | Publish the internal context version and next review date | Owner, version, date, and open questions |
Evaluate the pilot with five measures: fewer repeated instructions, fewer irrelevant results, fewer unsupported claims, more consistent buyer definitions, and faster human review. Do not judge the setup only by how much text the assistant produces. Record the approved SaleAI Customer Settings version beside the pilot results so later teams can reproduce the comparison.
The purpose of SaleAI Customer Settings is to give Copilot a reliable starting point while keeping task decisions and external evidence visible. Teams can explore SaleAI, review the SaleAI CRM, and request a walkthrough focused on Customer Settings, memory controls, field behavior, permissions, and the review process required by their business.
FAQ
What is SaleAI Customer Settings?
It is the visible configuration area for reusable company, product, market, customer, channel, and default email context used with SaleAI Copilot.
What information should be entered first?
Start with the official company identity, website, one priority product family, accurate applications, one target market, and the main customer types for the next pilot.
What is the difference between Read Memory and Allow Learning?
Read Memory brings saved context into the current conversation. Allow Learning controls whether the conversation may contribute to reusable memory. They should be reviewed as separate choices.
Should Read Memory always be enabled?
No. It is useful when approved context is relevant. Turn it off or compare results when the task is experimental, sensitive, unrelated, or intended to test whether saved context is causing an error.
Should Allow Learning always be enabled?
No. Keep it off for temporary assumptions, buyer-specific details, exploratory work, and unreviewed conclusions. Enable it only according to the team's approved memory policy.
How detailed should the product description be?
Include positioning, applications, important variants, buyer decision factors, available proof, and explicit exclusions. Avoid pasting an unstructured catalog or unsupported marketing claims.
Can SaleAI determine the correct HS code automatically?
The inspected Customer Settings contained an HS code field, but automatic legal classification was not verified. Treat any suggested code as a research aid and obtain qualified confirmation where required.
How many target markets should be saved?
Save only markets the team can support and review during the current operating cycle. Select one country or segment again in the task when precision matters.
Do email language and tone settings approve outreach?
No. They are drafting preferences. Recipient relevance, sender setup, message accuracy, compliance, and final activation still require review.
When should Customer Settings be updated?
Update them after material product, market, brand, certification, classification, or customer-strategy changes. Review stable fields quarterly and more dynamic preferences monthly.
