A field guide for the whole team

Meet Claude.
The colleague who's read everything and never gets tired.

You've heard the buzz. This guide cuts through it — a plain-English look at what Claude actually is, how it works, and the real ways it takes work off your plate at Younium. No hype, no jargon.

Show me how it works What's in it for my team?
~5 minto read the whole guide
0technical background required
Every teamCS, Support, Sales, Finance, Product & Dev
Start reading
01 — Start here

So what is Claude, really?

Claude is an AI assistant built by a company called Anthropic. The simplest way to think about it: you talk to it in plain language, and it talks back — helping you write, think, summarise, analyse, and get things done. Three mental models make it click.

A sharp colleague

A teammate who has read an enormous amount, writes quickly and clearly, never has a bad day, and is happy to do the first draft of almost anything. You still lead — they do the legwork.

A thinking tool, not a search box

Google finds links. Claude does the work with the information: it reads your messy notes, drafts the email, spots the risk in a contract, and explains its reasoning — in a back-and-forth conversation.

A very capable intern

Fast, eager, and broadly knowledgeable — but it can be confidently wrong, so you check its work on anything that matters. Give it good instructions and the results are remarkable.

02 — Under the hood

How it actually works

You don't need a computer science degree, but a basic picture helps you use it well — and trust it appropriately. Here's the whole thing in four steps.

1

It learned from a huge amount of text

Claude was trained on an enormous library of writing — books, articles, documentation, code. Through that, it absorbed patterns of how language, ideas, and reasoning fit together. It isn't a database it looks facts up in; it learned the patterns of how knowledge is expressed.

Deep reading, not memorising
2

It predicts the most useful next words

At its core, Claude predicts what should come next, one piece at a time, given everything you've said. That sounds simple, but at scale it produces genuine reasoning: it can plan, weigh options, write code, and explain itself. The quality depends heavily on what you give it.

Garbage in → garbage out
3

You give it context — it does the work

This is the part most people miss. Claude is only as good as the context you hand it. Paste in the transcript, the spreadsheet, the support ticket, the contract — then ask. The more relevant detail and clear instruction you provide, the sharper the result. A vague ask gets a vague answer.

Context is the magic ingredient
4

It can connect to your tools and act

Modern Claude isn't just a chat box. It reads your files, searches your documents, pulls from connected apps, builds spreadsheets and slide decks, and runs multi-step tasks on your behalf. That's the leap from "answers questions" to "gets work done."

From advisor to doer
03 — An honest scorecard

What it's great at — and where to be careful

Trust is built on honesty. Claude is genuinely powerful, but knowing its edges is what separates people who get great results from people who get burned.

Plays to its strengths

  • First drafts of anything — emails, docs, summaries, proposals. Beating the blank page is its superpower.
  • Summarising & extracting — turn a 60-minute transcript or a 40-page PDF into the key points in seconds.
  • Explaining & translating — make a complex topic simple, or rewrite something for a different audience or language.
  • Structured analysis — spot risks in a contract, compare options, organise messy notes into a clean table.
  • Writing code — from a spreadsheet formula to a full web page or presentation, no engineering background needed.
  • Repetitive thinking work — the tedious 80% of a task, leaving you the 20% that needs judgement.

Keep your hands on the wheel

  • It can be confidently wrong — it may state something false as if it's certain. Always verify facts, figures, and names.
  • It isn't live by default — unless it's searching or connected to a tool, treat its knowledge as a well-read generalist's, not today's news.
  • Sensitive data needs care — follow our company guidelines on what's OK to share. When unsure, ask.
  • It doesn't know our business unless you tell it — give it the Younium context and it gets dramatically better.
  • You own the final call — it's a powerful assistant, not a decision-maker. The judgement stays with you.
04 — Coming from ChatGPT

"We were using ChatGPT — why switch?"

Fair question — a lot of us asked it. ChatGPT is excellent, and everything you learned carries straight over. We're standardising on Claude because of how Younium actually works. Here's the honest, side-by-side.

Not a takedown. ChatGPT is strong and you weren't wrong to use it. This is about fit — for how we run (documents, tools, writing), Claude lined up better. On raw intelligence the two are roughly at parity, and both cost the same at the consumer tier. What separates them is where each one pulls ahead.
Where Claude is stronger

Why Younium standardised here

  • Writing codeIndependent tests and developer surveys consistently put Claude ahead on coding, and it's the default model in several popular developer tools.
  • Long & messy documentsA larger context window means full contracts, long transcripts and big threads stay coherent in a single pass.
  • Natural, professional writingProse reads more human and on-brand, with less of the formulaic feel — so it needs less editing.
  • Complex reasoning & instructionsStrong at following multi-part instructions and working through careful, step-by-step analysis.
  • Orchestrating across toolsChains steps across Jira, Slack, Outlook & Zendesk and keeps the whole investigation straight — not just firing off one action.
Where ChatGPT is stronger

Genuinely better — reach for it here

  • Generating images & videoCreates images and video natively. Claude doesn't generate images at all — for visuals, ChatGPT is the tool.
  • Real-time voice & audioA polished, conversational voice mode that Claude doesn't match.
  • Huge add-on ecosystemThousands of community-built GPTs and agents — an extensibility layer with no direct Claude equivalent.
  • All-in-one versatilityThe broadest single toolkit if you want every modality in one place.
05 — Closer to home

What this looks like at Younium

Abstract is easy to ignore. So here are concrete, real-world ways teams across Younium can put Claude to work this week. Pick your team.

Account health at a glance

Roll up recent emails, calls, and notes into a one-paragraph status on where each account really stands.

Account health

Prep for a check-in or QBR

Pull the account's history and goals into a briefing with talking points, wins to highlight, and risks to raise.

Check-ins

Keep long-term goals on track

Maintain each customer's success plan and flag what's slipping against the outcomes they signed up for.

Success planning

Spot renewal & churn risk early

Surface accounts going quiet or hitting friction so you can act before it becomes a renewal problem.

Retention

Draft ticket replies in seconds

Paste the ticket and the customer's history; get a clear, on-brand reply you can polish and send — instead of starting from scratch.

Tickets · Tone · Speed

Investigate the root cause

Feed it logs, error messages, or a confusing bug report and have it lay out the likely cause and the next steps to check.

Troubleshooting

Triage & prioritise tickets

Categorise a new ticket, set its priority, and route it to the right place — fast, and consistently.

Triage

Turn fixes into KB articles

A resolved ticket becomes a polished, reusable knowledge-base article so the same question never costs you an hour again.

Self-service

Prep for any call in 2 minutes

Drop in the prospect's notes and Claude builds a briefing: who they are, likely objections, and smart questions to ask.

Discovery

Answer RFPs & security questionnaires

Long vendor questionnaires drafted against our standard answers — a task that used to eat whole days.

RFP · Procurement

Tighten the follow-up email

Warm, specific, with clear next steps — written in the time it takes to grab a coffee.

Outreach

Decode complex billing asks

Translate a prospect's tangled pricing requirement into a clean summary of what they actually need.

Deal shaping

Make sense of messy data

Hand Claude an export and ask it to clean it, spot anomalies, or build the summary table you need.

Spreadsheets

Draft & document processes

Turn "how we actually do this" into a clean SOP or runbook the next person can follow.

Process · SOPs

Sanity-check billing logic

Walk through charge models — flat, quantity, volume, tiered — and pressure-test the edge cases before they bite.

Billing models

Flag the risks you might miss

Multi-currency, multi-entity, unusual contract terms — ask Claude to surface what deserves a second look.

Risk review

Synthesise user research

A pile of interview notes and support tickets becomes ranked themes and a clear set of recommendations.

Discovery

Write specs & tickets faster

Turn a rough idea into a structured PRD with goals, non-goals, and acceptance criteria.

Specs

Understand unfamiliar code

Paste a function and get a plain-English explanation, or draft tests and spot edge cases.

Engineering

Root-cause production issues

Trace an error across logs and recent changes to a likely cause and a recommended fix.

Debugging

Beat the blank page

Any document, email, plan, or message — get a solid first draft instantly and edit from there.

Universal

Catch up after time off

Summarise a long thread, a recorded meeting, or a busy channel so you're back up to speed in minutes.

Catch-up

Make hard things simple

Ask it to explain anything "like I'm new here" — the perfect patient teacher for unfamiliar topics.

Learning

Think out loud with a partner

Stress-test an idea, brainstorm options, or get the counter-arguments before you commit.

Thinking partner
Try this today — invoice template adjustments

Editing Younium invoice templates normally means carefully hand-writing Aspose LINQ Reporting Engine syntax. Hand the whole job to Cowork instead — here's the exact recipe.

STEP 01
Open it in Cowork

The desktop app that can read and edit your actual Word files — not just chat about them.

STEP 02
Give it the rules

Hand it the Younium references file (download below) — every field plus Aspose LINQ syntax. Or point it at the online docs.

STEP 03
Drop in the file + ask

Add the Word template and describe the change in plain language — "add a VAT line," "show the currency per charge."

STEP 04
Watch the magic

It hands back the adjusted template — correct syntax, null checks and charge-model logic handled for you.

The exact file Dennis uses — give it straight to Claude alongside your template.

The payoff: what used to be slow, error-prone hand-editing becomes a two-minute conversation — and you stay the reviewer who approves the result.

06 — The toolkit

The different ways to use Claude

"Claude" isn't one thing — it's a family of tools that share the same brain, each shaped for a different kind of work. Here's the lay of the land — the surfaces you'll actually open day to day.

Talk

Chat

The everyday conversation — the Claude app on web, desktop or mobile. Ask, draft, summarise, think out loud. Where almost everyone starts.

For everyone, every day
Do

Cowork

Works with your actual files and connected tools to carry out multi-step tasks — no coding. This whole guide was built in Cowork.

Desktop app only
Build

Code

Claude Code — a tool our engineers run from the terminal to delegate real coding: fixing bugs, writing features, navigating the codebase.

For developers
Design

Claude Design

Describe what you want and it builds the visual — slides, one-pagers, prototypes, web pages — then you refine by chatting. Export to Canva, PDF, PPTX or HTML.

Web only · claude.ai/design
Already live at Younium: we've built our own design system inside Claude Design. Paste in your text and it produces a fully on-brand Younium slide deck — laid out and formatted for you, automatically. No design skills, no fiddling with templates.
Why a lot of this works: Claude is genuinely excellent at writing code

You don't have to be technical to benefit. The clearest example: we're shifting from PowerPoint to HTML presentations. Claude builds them straight from plain text, they look sharper than a templated deck, open in any browser, and share as a simple link — this very guide is one. PowerPoint still works whenever you need it; Claude happily does both.

07 — Power features

The features that quietly save hours

Beyond the chat box sit a handful of capabilities that turn Claude from a smart assistant into something that genuinely takes work off your plate. These are the ones worth knowing.

Skills

Remember Neo in The Matrix — "I know kung fu" the second it's uploaded? Skills are exactly that: installable expertise that makes Claude an instant specialist at a specific Younium job — answering RFPs from our knowledge base, adjusting invoice templates, running a support investigation, building branded slides. Claude picks the right skill automatically; you just describe the task.

Neo from The Matrix waking up after instantly learning kung fu

"I know kung fu." — basically how it feels.

Scheduled tasks

Set Claude to run work on a schedule — a Monday-morning summary of your week, a daily check on something you're watching, a monthly Jira report. It does the work and the result is waiting for you, without you having to remember to ask.

Scheduled tasks running on their own
How to add a skill. Many Younium skills are already set up for you — Claude just uses them. To add more, open the Skills (or Plugins) area in Claude's settings and install one from the marketplace, or upload a skill file someone's shared. Once it's in, you never "launch" it — Claude picks the right one automatically when your request matches. Need a skill that doesn't exist yet? Ask Dennis.

Make Claude respond your way

You can give Claude standing instructions for how to respond and act — tone, format, the terminology it uses, even the role it should take. Set them once and every answer arrives shaped that way, so you stop repeating yourself in each chat. One of the most useful: tell it not to make assumptions but to verify, and to ask you when it's unsure — so a vague prompt comes back with a few sharp questions to guide you, instead of a confident guess.

A real Younium example

"When I share invoice-template questions, use Aspose LINQ Reporting Engine syntax with proper null checks and charge-model branching. Use Younium terminology. Be concise and don't hedge. Don't assume — verify, and if anything is unclear, ask me before you answer."

Set that once and every relevant reply follows it — no re-explaining. With that last line in place, a one-line prompt comes back as a couple of pointed questions that steer you to the right answer. You can keep personal instructions, and teams can share a set so everyone gets consistent results.

08 — Models & usage

Pick a model — then mind the context

The last piece: which Claude you're talking to, and how usage works. Two simple ideas explain almost everything about speed, cost, and limits — Claude reads in tokens, and it re-reads the whole conversation every time you reply.

One brain, three speeds

Claude comes as three models — same family, different balance of speed and depth. They're improved constantly, so whatever the task, pick the newest version available for the best results.

Fastest

Haiku

The quickest and lightest. Ideal for simple, high-volume, or quick-turnaround tasks where speed matters most.

Quick, simple tasks
Balanced

Sonnet

Fast and highly capable — the everyday workhorse, and the right default for the vast majority of work.

Your day-to-day default
Deepest

Opus

The most capable, for the hardest thinking — complex analysis, tricky code, long multi-step reasoning. Slower, but the heavyweight.

Complex, high-stakes work
Don't bother memorising version numbers. New, smarter versions ship regularly — just grab the latest one on offer and you're getting the best Claude available.

Context, tokens & limits

Now the cost side. The same two ideas drive your speed and limits: Claude reads in tokens, and it re-reads the whole conversation on every turn.

Tokens

The unit Claude reads and writes in. A token is roughly ¾ of a word — "subscription" is about three. Everything you send and everything it replies is counted this way.

The context window

Claude's working memory for a single chat. Your files, the back-and-forth, its own replies — all of it has to fit inside one window. Large, but not infinite.

It re-reads every turn

The part most people miss: each time you hit send, Claude re-reads the entire conversation so far. The longer the chat, the more it processes for every new reply.

Watch one conversation grow

Keep replying in the same chat and the window fills — and every new answer re-reads everything above it. Add a few turns and watch the cost of each reply climb.

Context window
getting full
Turn 1
Conversation so far1,700 tokens
This reply re-reads1× the first message's work
Cost per reply

Long chats get slower and pricier per reply. Switching topics? Start a fresh chat — it's cheaper, and sharper too.

Your usage budget — and its limits

Access works like a budget that refills on its own. Two limits run at once: a rolling 5-hour window and a weekly cap. Heavier models and longer chats draw it down faster. Pick a model and run a few tasks to see how it behaves.

5-hour window100%
Refills on a rolling 5-hour basis
This week100%
Resets at the start of each week

Running tasks on Haiku — the lightest draw on your budget.

In practice, you'll rarely hit a wall. Day-to-day work sits comfortably inside the limits, and anything you do hit simply resets. The habits that help: start a new chat for a new topic, keep the right model for the job, and don't paste in more than the task needs.
09 — The superpower

Connect your tools — and find the root cause in one place

This is the leap a plain chatbot can't make. Connect Claude to the systems where your work actually lives, and it can investigate across all of them at once — instead of you opening ten tabs and copy-pasting between them.

Jira logo Jira
Slack logo Slack
Outlook logo Outlook
Zendesk logo Zendesk
＋ your files & more

A real investigation, start to finish

The scenario: a customer reports their invoices stopped syncing on Tuesday. Watch Claude run the trace across every system at once.

1

Start in Zendesk

Reads the ticket itself, the customer's history, and exactly what they reported — so the investigation starts from the real symptom, not a guess.

The symptom
2

Cross-check Slack

Searches channels for anyone already discussing this customer or this symptom — maybe someone hit it on Tuesday and there's a thread you never saw.

Has anyone seen this?
3

Check Jira

Looks for a known bug, a linked ticket, or a recent deploy that lines up with when it broke — connecting the customer's pain to an engineering cause.

The likely cause
4

Scan Outlook

Pulls the email thread with the account so nothing already promised to the customer gets missed or contradicted.

What we've told them
5

Hand you the verdict

A plain-English root-cause summary, the evidence behind it, and a recommended next step — assembled from four systems, in one place, in the time it takes to read this.

Decision-ready
What you'd actually type
"Customer Acme says invoices stopped syncing since Tuesday. Check the Zendesk ticket, search Slack and Jira for anything related, and the Outlook thread with their CSM — then tell me the likely cause and what to do next."

→ What comes back: the matching Zendesk ticket, a Slack thread from Tuesday, a Jira bug about the sync job, the customer's email history — and a one-paragraph verdict with a clear recommended action.

Connecting a tool is a one-time setup, and access always follows our data-handling guidelines — Claude only sees what we deliberately connect.

When you just need an answer: Ask Younium

Not every question needs a full investigation. Ask Younium is Claude pointed at our own knowledge — handbooks, processes, product docs, past answers — so anyone can get a reliable, sourced answer to "how do we do this here?" without pinging a colleague or digging through Confluence.

How do we handle a mid-term upgrade with proration on an annual contract?

Younium

Add the new charge with an effective date mid-term — Younium prorates it against the remaining period automatically, and the next invoice picks up the delta. Keep the original term-end date so the renewal logic stays intact.

Sourced from Subscription handbook Billing FAQ
The difference from a web search: answers come from our documented way of working, with the sources attached — so new joiners ramp faster and everyone answers customers the same way. It only knows what we've connected, and it always shows its sources so you can verify.
10 — Before & after

The difference, in real terms

Forget the theory for a second. Here's what a normal Younium task looks like with and without Claude — the same job, two very different afternoons.

Invoice template tweaks

Before

Hand-writing Aspose LINQ syntax, trial and error, double-checking null cases — a careful afternoon gone.

After

Drop in the Word file, describe the change in plain English, get the corrected template back to review.

An afternoon → a few minutes

A support root-cause

Before

Ten tabs open across Zendesk, Slack, Jira and Outlook, piecing the story together by hand.

After

One prompt. Claude traces it across all four and hands back the cause, the Jira ticket, and a drafted reply.

30+ minutes → seconds

RFP & security questionnaires

Before

Days of copy-pasting from old responses, hunting for the right wording, hoping nothing's missed.

After

Drafted against our knowledge base in one pass, with the uncertain items flagged for a human to confirm.

Days → about an hour

Meeting follow-ups

Before

Re-reading scratchy notes after the call to reconstruct decisions, actions and who owns what.

After

Paste the transcript — get decisions, action items and owners as a clean list, instantly.

20 minutes → instant

On-brand slide decks

Before

Wrestling PowerPoint templates, nudging boxes, fixing fonts so it looks properly Younium.

After

Paste your text into our Claude Design system — a fully on-brand deck comes back formatted.

Half a day → minutes

Account check-in prep

Before

Digging back through emails, notes and calls to remember where the account actually stands.

After

A one-paragraph status with talking points and risks, ready before you join the call.

20 minutes → 1 minute
11 — Your first 10 minutes

Getting started

The hardest part is the habit, not the tool. Here's how to go from curious to useful today.

  1. Install the app, then start a conversation

    Grab the Claude desktop app (download button at the bottom of this page), sign in, and type a real task you're facing right now — in normal language, like you'd brief a colleague.

  2. Give it the context

    Paste in the email, the notes, the data, the ticket. The more it has to work with, the better the answer.

  3. Be specific about what you want

    Say the format, length, tone, and audience. "A short, friendly reply to this customer" beats "help with this."

  4. Iterate — don't accept the first try

    Reply with "make it shorter," "more formal," "you missed X." It's a conversation, not a vending machine.

  5. Check anything that matters

    Verify facts and figures before they go out the door. You're the editor; Claude is the drafter.

The single biggest tip: be specific
"Write an email."
↳ "Draft a warm but professional email to a customer who's frustrated their invoice ran in the wrong currency. Apologise, explain we're fixing it, and give a clear next step. Keep it under 120 words."
"Summarise this."
↳ "Summarise this meeting transcript into: decisions made, action items with owners, and open questions. Use bullet points."

Notice the pattern: who it's for, what format, what to include, how long. That's 90% of getting great output.

12 — The elephant in the room

The questions everyone's thinking

Let's address the hesitations honestly — because the people who haven't started usually have one of these on their mind.

The realistic answer: Claude replaces tasks, not people — and mostly the tedious ones. It handles the first draft, the summary, the data cleanup, so you spend your time on judgement, relationships, and the decisions only a human should make. The people who thrive are the ones who learn to delegate the busywork to it and focus on higher-value work.

Trust it like a fast, well-read colleague who occasionally misremembers — useful and often right, but worth checking on anything that matters. It can state wrong things confidently. For drafts, brainstorming, and summaries of text you gave it, it's extremely reliable. For facts, figures, names, and dates, verify before you act.

Good instinct to ask. Always follow our internal guidelines on what data is appropriate to share, and when in doubt, check with the relevant person before pasting anything sensitive. Used within our guidelines, it's a safe and powerful tool — but the responsibility to handle data correctly stays with all of us.

Especially for you. There's nothing to install or code. If you can write a message to a colleague, you can use Claude. In fact, non-technical teams often get the biggest wins, because so much of their work is writing, summarising, and organising — exactly what it's best at.

Almost always, the issue is the input, not the tool. A vague, one-line request gets a generic answer. Give Claude real context and a specific ask — like the examples above — and the difference is night and day. Give it ten honest minutes with a real task and judge it on that.

The best way to understand Claude is to use it on something real.

Pick one task from your week that you've been dreading. Hand it to Claude. That's the whole pitch.

Try it on a real task Download the desktop app

Don't have access to Claude yet? Ask Everart.