Topic
AI
-
What is AI?
What the word covers, what changed recently, and why AI and ChatGPT are not the same thing.
-
Analyze a spreadsheet by asking questions in plain English
Useful findings, anomalies, and follow-up questions from a spreadsheet, without needing to know formulas or pivot tables.
-
Before you pilot Claudeforce
Connecting Claude to your org doesn't loosen your security model, but it means your existing permissions matter more than they used to. Here's what to check first.
-
Bots, chatbots, assistants, and agents
Four words that get used as if they mean the same thing, and one customer example that shows where each one stops.
-
Can AI be wrong?
Yes, and in the same confident tone it uses when it's right. Where that matters most, and how to work with it anyway.
-
Claudeforce versus Agentforce, in plain terms
If you already use Agentforce, Claude's been running inside it since late 2025. What Claudeforce actually adds is mostly administrative, not algorithmic.
-
Company, product, or model?
OpenAI, ChatGPT, and GPT are three different things. What each layer is, and why the names are legitimately confusing.
-
Compare a few options and help me see the differences
An apples-to-apples comparison, with tradeoffs surfaced and the unanswered questions named, without letting it make the choice for you.
-
Compare vendor proposals without missing the fine print
A normalized comparison across vendor proposals, plus the gaps and questions each one still leaves open.
-
Create a FAQ from the questions customers keep asking
A draft FAQ organized by what customers are actually trying to do, with the approved facts kept separate from the draft wording.
-
Describe your experience for a job application
Turn what you've actually done into clearer wording for a resume or application, without inventing a thing.
-
Documenting a validation rule so the next person understands it
The formula says what a validation rule blocks. It doesn't say why. Here's how to turn it into a description someone can act on without opening the builder.
-
Explain something technical in normal language
Translate jargon into plain language you can act on, without losing the one detail that changes what you should do.
-
Explaining CPQ logic in plain language
A CPQ price rule is unreadable to anyone who didn't build it. Here's how to get it translated for a non-technical approver, and what to verify before trusting the translation.
-
Follow up on money a customer owes you
A clear, professional reminder built from the facts you have, without threats or fees you never agreed on.
-
I got a letter or document I don't understand
A plain-language explanation of what a confusing letter actually says, the parts that matter, and what to verify before you act on it.
-
Is AI searching the internet?
Sometimes. A fluent answer doesn't mean anything was looked up, and here's how to tell the difference.
-
My bill went up. What changed?
Compare two bills side by side, find exactly what changed, and go into the phone call with the right questions.
-
Plan a project around the house
Turn an idea into a materials list, an order of steps, and the questions to answer before you buy anything.
-
Plan a trip around my actual preferences
A realistic outline built around your actual dates, budget, and pace, not a generic top-ten list.
-
Prepare questions for an appointment or important conversation
An organized question list, priorities, and the facts you don't want to forget, before a conversation you only get one shot at.
-
Research something without opening twenty tabs
A research brief built for a specific decision, with claims kept separate from what still needs checking.
-
Respond to a bad review
A calm public reply, based only on what you know happened, that future customers will read as fair.
-
Reviewing SOQL for readability and risk before it ships
A query that runs fine on ten test records can lock up at scale. A checklist for having AI review SOQL for risk, and where to double-check it yourself.
-
Summarize a long document and find what needs your attention
A short summary of a long document, the parts that ask something of you, and a check for what got left out.
-
This school email is long. What do I actually need to do?
Pull the actions, dates, and requirements out of a long school email, so nothing gets lost under the announcements.
-
Troubleshoot something from a photo
Likely causes, safe checks, and what to gather next, from a photo of something broken around the house.
-
Turn customer feedback into themes
A concise set of themes from feedback or reviews, each with representative examples and possible actions.
-
Turn meeting notes into decisions and action items
A decision log, owners, due dates, and unresolved questions, with anything inferred flagged instead of presented as fact.
-
Turn messy notes into a useful plan
A clean summary, the decisions already made, the open questions, and the next actions, straight from your messy notes.
-
Turn the way I do something into an SOP
A repeatable procedure someone else can follow, built from a rough explanation and an interview for the steps you left out.
-
Turning business requirements into Salesforce stories
A requirements doc and a story a developer can actually build from are different documents. Here's how to get from one to the other without losing the original ask.
-
What can AI see and remember?
It depends on the product and what you've turned on. What it usually sees, what memory means, and where to check yours.
-
What Claudeforce actually is
Salesforce and Anthropic announced Claudeforce in August 2026. Here's what the two halves actually do, and what's confirmed versus still vague.
-
What Claudeforce can't do yet
Beyond the press release, there are real open questions here: unquantified cost, a beta colliding with a Salesforce release window, and a vendor concentration risk architects are already flagging.
-
What do I actually type or say to AI?
You don't need special wording. Four parts that turn a vague question into a useful request.
-
What is a prompt?
The word for what you give an AI to work with, and why it doesn't need to be clever.
-
What should I never share with AI?
The short list to keep out, and a worked example of removing personal details before you ask for help.
-
What to do when AI isn't allowed to see the real data
The records are locked down, but most of the work around them isn't. A way to keep going with sample data instead, and how to know when a sample won't hold up.
-
What to do when the connector says it worked but nothing happened
It said the action succeeded. The system it was supposed to change looks untouched. Here's where that gap usually comes from, and how to find it.
-
Write a job posting
Turn your rough idea of the role into a clear posting, without requirements or perks you never mentioned.
-
Write a proposal from rough notes
A structured proposal draft with scope, assumptions, and next steps, built from a rough description of the job.
-
Write the email or message you've been putting off
A sendable draft, in the tone you actually want, for the message you keep opening and closing without sending.
-
Writing a Flow description AI can actually follow
AI can't open your Flow. Here's a plain-text way to describe one that's specific enough to actually get help with it.
-
The Claude certifications, and whether you can actually take one
Four exams exist. Most people reading this are not eligible to sit them, which is worth knowing before you spend a weekend studying.
-
CCAO-F: the Claude exam for people who don't build things
The associate exam is the only one of the four written for non-technical roles. It is also the only one that does not count toward your employer's partner status.
-
CCAR-F: the architect exam that tests what you normally look up
Sixty questions built from six published scenarios, and the most detail-heavy exam of the four. "Foundations" describes the level of detail, not the difficulty.
-
CCAR-P: the exam that asks you to justify, not to configure
The only Professional-level exam of the four, and the most expensive. Its objectives use verbs like evaluate and justify, and two of seven domains are about talking to people.
-
CCDV-F: the developer exam, and the two domains not to overstudy
One domain is a third of the exam. Two others are worth about three questions between them. This is the only one of the four that publishes a weight for every skill.
-
Argue me out of it
You've mostly decided. Find the questions you haven't asked before you commit.
-
Read this before you sign it
Use AI to understand a lease, a contract, or a policy, and know exactly where it will let you down.
-
The letter you've been putting off
Disputing a charge, canceling something that resists canceling, saying no to someone you'll see again.
-
How to add an MCP server to Claude Desktop
The one-click way, then the way that works when the thing you want isn't in the list.
-
Your first useful hour with AI
Three small tasks to try before you decide whether any of this is worth your time.
-
Build your own MCP server
For when nothing in the directory does the job. This is the one guide here that asks you to open a terminal.
-
How to check its work
It writes wrong answers in the same confident tone as right ones. Three ways to tell them apart.
-
Stop retyping yourself
You type the same five facts about your job every day. There are two places to put them instead, and a rule for which goes where.
-
What you should never paste, and why
A short, clear line between what's fine to share and what could get you in trouble, plus what to do if you already pasted it.
-
Why it forgot what you told it
Long conversations quietly get worse. The tells, and how to move to a fresh one without losing the work.
-
Why your answer was disappointing, and the four fixes
The output was vague because the input was. The four things missing, in the order they matter.
-
How to use AI without connecting it to your Salesforce org
Security said no. Requirements, logic design, and specs were never the blocked part. Here is the four-step workflow, with the prompts.