Responsible use of AI
Where AI tools may help our delivery and content, where they do not, and the rules we follow so your data and your decisions stay protected.
AI tools can speed up parts of consulting work: drafting documentation, suggesting test cases, explaining unfamiliar code. They can also produce confident, wrong answers about Salesforce behaviour, and they create real risk if client data is pasted into the wrong place. This page explains how GeekMidst approaches both sides.
It covers two things: how we may use AI in our own delivery and content, and how we approach AI features you ask us to build, such as Agentforce. It is a statement of practice, not a legal document; your contract and your own AI policies take precedence.
Principles we follow
- A person is accountable for every output. AI may help draft, but a named GeekMidst consultant reviews, corrects and takes responsibility for anything we deliver.
- No client data in public AI tools without consent. We do not paste client data, org metadata containing personal data, credentials or confidential documents into public AI services unless you have explicitly agreed, in writing, to a specific tool and use.
- Your policy comes first. If your organisation bans or restricts AI tools, we follow that policy on your engagement.
- Verify against the platform. AI suggestions about limits, behaviour or syntax are checked against Salesforce documentation and tested in a sandbox before we rely on them.
- Be open about it. If you ask whether AI was used in producing a deliverable, we will tell you.
Where AI may and may not be used
| Use of AI | Conditions | |
|---|---|---|
| Drafting documentation and runbooks | Possible | Using non-confidential structure only; reviewed and completed by a consultant |
| Suggesting code, SQL or AMPscript | Possible | No client data in prompts; code reviewed, tested in a sandbox and owned by a developer |
| Generating test scenarios | Possible | Scenarios reviewed against real requirements; test data is synthetic |
| Analysing client data or exports | Only with written consent | Approved tool, agreed scope, agreed retention |
| Making design or go-live decisions | No | Decisions are made by people, with you |
| Website articles and guides | Possible as a drafting aid | Technical claims checked by a consultant before publishing |
When you ask us to build AI features
Many organisations want AI inside their own Salesforce solutions, for example Agentforce agents, generated email copy or AI-assisted case handling. Through our Agentforce and AI automation work we apply the same thinking:
- Start with the data. AI features are only as good as the records, permissions and grounding data behind them. We review data quality and access before design.
- Define boundaries. Agents and prompts get clear scopes, the actions they can take and the cases where a human must take over.
- Respect consent and access. AI features should not expose data a user could not otherwise see, or use personal data in ways customers have not agreed to. Review how the Einstein Trust Layer and your org settings apply; check Salesforce Help for current capabilities in your edition.
- Test like any other change. That includes unexpected inputs, edge cases and attempts to push the feature outside its scope.
- Keep humans in the loop for decisions that significantly affect customers.
Content on this website
Our guides may be drafted with AI assistance. Each is reviewed by a person for technical accuracy before it is published. If you spot an error, tell us at info@geekmidst.com and we will correct it.
Related pages
Security and data handling
How we access your orgs and treat your data.
→Agentforce and AI automation
Designing and testing AI features in Salesforce.
→Our approach
The phases and quality gates every engagement follows.
→Privacy policy
How this website handles personal data.
→Questions we hear
Will you use AI on our project?
Can we approve a specific AI tool for use on our data?
Who is responsible if AI-assisted code has a bug?
Do you train AI models on client data?
Planning AI features in Salesforce?
If you are weighing Agentforce or other AI capabilities, we can help you assess data readiness, scope and risks before you build.
