Why Your Team May Stop Opening Business Apps
You may recognise this routine: a salesperson opens the customer database, searches several records, checks the latest messages, reviews outstanding tasks, updates a deal stage and then prepares for a meeting. None of those steps is especially difficult, but together they consume attention that could have gone into serving customers.
For a small business, this problem is even more visible. You may have one person handling sales, follow-ups and administration, while another manages customer support and reporting. When information is spread across a CRM, email, spreadsheets and messaging apps, your team spends too much time finding context before taking action.
Salesforce and Anthropic are testing a different approach: placing Salesforce data and actions inside Claude, Anthropic’s AI assistant. The announced Salesforce in Claude plugin includes 37 pre-built sales skills and allows users to query, update and act on live CRM information without opening the Salesforce application. The feature is currently being tested with selected customers, with an open beta planned for September, according to VentureBeat.
TL;DR
AI assistants are becoming a new way to use business systems, rather than merely tools that sit beside them.
For your SME, the important lesson is to organise customer data, permissions and workflows so an assistant can help your team act safely and consistently.
What This Means
The idea is simple. Instead of asking employees to navigate menus and search through records, they can ask Claude a direct question such as, “Which opportunities need attention this week?” or “Prepare me for my meeting with this customer.” The assistant then reads the relevant CRM information, follows defined instructions and returns a useful summary.
It may also perform an approved action, such as updating a deal record or creating a follow-up task. The Salesforce plugin uses Salesforce’s existing permissions through its Model Context Protocol, or MCP, connection. In plain language, a user should only be able to view or change records that the user already has permission to access, according to the explanation given by Salesforce to VentureBeat.
This is different from a chatbot that simply generates text. The assistant is connected to your operational data and business rules. It can reason over records, identify the next step and, where authorised, carry out an update.
The key lesson is not that your team needs another app. It is that your existing business information should become easier to use without weakening control.
Salesforce describes this as a headless approach: the CRM’s data, workflows and governance can be used without relying on its traditional human-facing screen. Salesforce released Headless 360, including APIs, MCP servers and command-line tools, before turning the concept into a simpler Claude plugin, as reported by VentureBeat.
How This Applies to Malaysian SMEs
Consider a local distributor with a small sales team serving retailers across Selangor, Johor and Penang. Before a weekly sales meeting, the owner may ask each salesperson to prepare updates manually. A connected assistant could instead summarise open opportunities, identify records without recent activity and list customers waiting for quotations. The owner receives a more consistent view, while the salesperson spends less time assembling a report.
For a service business, the same approach could help with customer follow-up. Imagine an air-conditioning maintenance company whose staff record enquiries, site visits and quotations in a CRM. A team member could ask, “Which customers received a quotation more than seven days ago and have not replied?” The assistant could produce the list and prepare follow-up tasks. The business still needs a person to review the suggested action, but the searching and sorting becomes much faster.
A Malaysian wholesaler or importer could use the concept to support account management. A manager might ask for customers whose order frequency has fallen, accounts with unresolved complaints or buyers whose promised payment date has passed. The assistant would not replace the manager’s judgement. It would bring relevant information together so the manager can decide what to do next.
The approach is also relevant to businesses that do not use Salesforce. You may operate with a smaller CRM, accounting platform, helpdesk, inventory system or structured spreadsheet. The principle remains the same: when your records are consistent and your processes are clearly defined, automation can help employees find information and complete routine work through natural language.
However, you should not connect an assistant to every system immediately. Start with one process where delays are easy to see, such as lead follow-up, quotation tracking, appointment preparation or customer issue escalation. This gives you a manageable test and makes it easier to check whether the assistant is producing reliable results.
What You Should Prepare Before Using This Approach
The quality of an AI assistant depends heavily on the quality of the information it can access. If customer names are duplicated, deal stages are inconsistent or important conversations remain in personal messaging accounts, the assistant may produce incomplete answers.
You should also define what an assistant may do automatically and what requires human approval. Reading a customer record and drafting a follow-up message may be low risk. Changing a quotation status, deleting a record or sending a message to a major customer deserves stronger controls.
Permissions matter as well. Review access by role rather than giving every employee broad visibility. A salesperson may need access to their own opportunities, while a manager may need team-level reporting. Your assistant should follow the same boundaries.
A Simple Readiness Checklist
- Clean your records: standardise customer names, phone numbers, sales stages and follow-up dates.
- Choose one workflow: begin with a task such as meeting preparation or overdue enquiry follow-up.
- Write clear rules: define when a task should be created, escalated or reviewed.
- Set user permissions: confirm which roles can view, edit or approve each type of information.
- Require approval for sensitive actions: especially customer communications, status changes and financial records.
- Test with real examples: compare the assistant’s result against a manually checked answer.
- Keep an audit trail: record what the assistant recommended or changed.
Useful Numbers to Track
The Salesforce and Anthropic announcement provides several useful indicators of the direction of this technology. Treat them as signals about product development, not promises about your own results.
| Data point | What it tells you | Source |
|---|---|---|
| 37 pre-built sales skills | The initial focus is on repeatable sales tasks such as meeting preparation and pipeline analysis. | VentureBeat |
| 30 seconds | Salesforce described a potential reduction in the time needed to synthesise opportunity information, compared with extensive manual navigation. | VentureBeat |
| 10,000 clicks | Salesforce used this figure to illustrate the amount of interface interaction it believes can be removed from a seller’s morning review process. | VentureBeat |
| September open beta | The product announcement indicated that broader testing was planned after the selected-customer pilot. | VentureBeat |
Practical Takeaways
- Do not judge automation by whether employees stop opening an application. Judge it by whether work is completed accurately and on time.
- Make your CRM or customer database the trusted source of record before connecting an AI assistant.
- Document your sales and service processes in plain language so the assistant has clear instructions to follow.
- Start with information retrieval and draft preparation before allowing automatic updates.
- Ask staff to check AI-generated summaries against the original records, especially for important customers.
- Use role-based access so confidential customer, employee and business information is not exposed unnecessarily.
- Measure follow-up completion, response time, record accuracy and manager visibility after introducing the workflow.
The Bigger Picture
The long-term change is that business software may become less dependent on fixed screens. Employees will still use dashboards when they need a visual overview, but many routine tasks may begin with a question or instruction.
For SME owners, this creates both an opportunity and a responsibility. The opportunity is to help a small team perform work that previously required more administrative effort. The responsibility is to ensure that the underlying data, permissions and approval rules are dependable.
This trend also means you should think carefully about where your business knowledge lives. If your best sales practices exist only in one experienced employee’s memory, an assistant cannot reliably apply them. If follow-up rules, customer categories and service procedures are written down and reflected in your systems, they can become repeatable.
Salesforce’s partnership with Anthropic shows that major software platforms are willing to let users access their capabilities through an external AI interface. The practical message for your business is straightforward: do not focus only on adopting the newest assistant. Focus on building organised information and clear workflows that any trusted assistant can use safely.
Ready to Streamline Your Operations?
Your business should run itself. AutoRunBiz deploys AI agents to automate your daily operations — WhatsApp orders, invoicing, customer follow-ups, and accounting. Book a free 15-min ops audit to see where automation fits your business →