In plain English
- Markdown: plain text with a few symbols for headings and lists, so any device or app can open it.
- Handover: a note that tells a new chat, or the next person's AI, what was done and what comes next.
- Spec: short for specification, a plain description of exactly what a piece of output must contain.
- Daily log: a short note of what you did today, one per day, written by your AI.
- Project: a space in Claude or ChatGPT that keeps files and instructions for every chat inside it.
On 13 November I was about to move from my laptop to my desktop, so I asked Claude for a handover prompt that would let me "pick this right up". What came back was a spec: where my draft and my notes were, how to check where the work stood, the state of each piece, the rules for the work, then the next steps in order. I pasted it into a new session on the other machine and the work carried on from it.
That handover is the kind of file I covered in Ask your AI to save its work as files, where I wrote about keeping a daily log and asking for a handover doc. This second part is about how to write them for a new chat, the next step in a job, or a teammate's AI. We have to think about how a computer wants to work with it rather than how a human wants to work with it. For an AI reading your handover, that means a short list of where things are and what to do next.
Ask for a spec at the end of every chat: "a defined spec to output", which means a list of exactly what the handover must contain. Skip pasting the whole chat: it's long, and most of it is back-and-forth the next chat doesn't need. Skip the Word document some teams keep, full of links to other chats, too: that's "creating extra steps for not just the bot but for yourself". In Claude or ChatGPT, paste this request in at the end of a chat.
Write a handover for my next chat. Don't paste our chat; write only what a new chat needs. Start with a line on what this is, then these headings: Where things are, Current state, Rules, Next steps (in order).
To decide what goes in the format, ask what the next step needs in order to do a good job, then work backwards from there. These four steps keep every handover on the same headings.
Next chat: paste the handover that comes back into your next chat.
When a handover comes out well: ask the AI to turn its headings into a format you can reuse.
Write the headings from that handover as a format I can reuse, with one line under each heading saying what goes there.
Keep the format with your work: save the reply and add it to your project yourself. In Claude it goes in the project's knowledge, and in ChatGPT in the project's files.
From then on: end chats in that project with this request.
Write the handover using the handover format in this project.
If you work in folders with Claude Code or Claude Cowork, have the AI save the format as a Markdown file in your project folder instead. It can then write each handover, and your daily log, straight into that folder, or into a folder your team shares.
In one test on a made-up chat about choosing an office coffee supplier, a new chat given only the handover chose the same supplier as a new chat given the whole chat. The handover was 440 words and the chat was 1,598.
Real replies from Claude Sonnet 5.5 in Claude Code on 15 Nov 2026, each prompt run once. No AI is running on this page.
Whole chat or handover
A made-up chat of six messages and six replies about choosing an office coffee supplier was handed to a new chat twice: once as the whole chat, once as only the handover. Both new chats then got the same message.
The message both new chats got
Supplier C has confirmed they deliver in 3 days. Carry on from where we left off.
The new chat got the whole chat, all six messages and replies: 1,598 words.
Show the whole chat it was given
You
I need to choose a new coffee supplier for our office of 12 people. Help me compare three quotes. Supplier A: £18 per kilo, delivers in 2 days, minimum order 5kg. Supplier B: £15 per kilo, delivers in 5 days, minimum order 10kg. Supplier C: £16.50 per kilo, delivery time not confirmed yet, minimum order 3kg.
Claude's reply
Coffee supplier comparison
Assumptions
I assumed about 2 cups per person per day at roughly 10g of coffee per cup. For 12 people on about 21 working days a month, that is about 5 kg/month, or 60 kg/year. If your usage differs, the totals below scale accordingly.
Cost
| A | B | C | |
|---|---|---|---|
| Price per kg | £18.00 | £15.00 | £16.50 |
| Minimum order | 5 kg (£90) | 10 kg (£150) | 3 kg (£49.50) |
| Delivery | 2 days | 5 days | Unconfirmed |
| Cost at 5 kg/month | £90 | £75 | £82.50 |
| Cost at 60 kg/year | £1,080 | £900 | £990 |
| Saving vs A | – | £180 | £90 |
Trade-offs
- A is the most expensive, but it is fast and its minimum order matches your monthly usage. That makes it the easiest to manage and the freshest to drink. It also has the lowest risk of running out.
- B is the cheapest per kilo, but a 10 kg minimum is about two months of supply. Roasted beans are best within roughly 4-6 weeks of the roast date, so the second half of each order will taste staler. It also needs storage space and ties up cash. The 5-day delivery means you have to reorder early.
- C is mid-priced and has the most flexible minimum, so you can order small and adjust. Its only problem is the unconfirmed delivery time, which makes it impossible to rank properly yet.
Recommendation
C looks like the best balance if it can deliver in about 5 days or less. It is £90/year cheaper than A, and the 3 kg minimum avoids the stale-stock problem. If C's delivery turns out to be slow or unreliable, pick A for reliability. I'd only pick B if you can store coffee well, for example by splitting it into airtight portions, and the £180/year saving matters.
Questions to ask before deciding
- C: What is the delivery time?
- All three: Is delivery free, or is there a shipping charge? This could change the ranking, especially for small orders.
- All three: Is the roast date guaranteed (for example, roasted to order)?
- All three: Are the quotes for whole beans or ground, and is it the same blend and quality?
- All three: Is there a contract or subscription lock-in, and do they offer discounts for larger or repeat orders?
- All three: Can you get a sample bag first? Taste matters more than a £90/year difference.
If you tell me your actual monthly usage and whether delivery is charged, I can redo the numbers.
You
Actually, forget the minimum order. We have plenty of storage. Compare on price and delivery only.
Claude's reply
Price and delivery only
| Price/kg | Delivery | Annual cost (60 kg) | vs cheapest | |
|---|---|---|---|---|
| B | £15.00 | 5 days | £900 | – |
| C | £16.50 | Unconfirmed | £990 | +£90 |
| A | £18.00 | 2 days | £1,080 | +£180 |
The annual figures use my earlier estimate of about 5 kg/month. Without the minimum orders, the ranking doesn't depend on volume. B is always £1.50/kg cheaper than C and £3.00/kg cheaper than A.
- Price: B is cheapest, then C, then A.
- Delivery: A is fastest at 2 days, then B at 5 days. C can't be ranked until it confirms a date.
Takeaway
- B is the best choice if you can plan ahead. With plenty of storage, the 5-day delivery just means you reorder before you run low. It saves £180/year over A.
- A costs £180/year more than B, and the only thing it buys you is 3 fewer days of waiting. That's only worth it if you tend to run out unexpectedly.
- C sits in the middle on price. Its delivery time decides whether it beats B, but it would need to be much faster than 5 days to justify an extra £90/year.
Ask C for its delivery time. If it's 5 days or more, B wins outright. If it's 2-3 days, it's a reasonable middle option.
Since you're no longer weighing freshness against storage, ask all three about roast dates. Large, infrequent orders can mean older beans.
You
We get through about 4kg a month. What would each one cost us per month?
Claude's reply
At 4 kg a month:
| Price/kg | Monthly cost | Annual cost (48 kg) | |
|---|---|---|---|
| B | £15.00 | £60.00 | £720 |
| C | £16.50 | £66.00 | £792 |
| A | £18.00 | £72.00 | £864 |
B is £6/month cheaper than C and £12/month cheaper than A. Over a year that's £72 versus C and £144 versus A.
These figures are for the coffee only and ignore delivery charges. If any supplier charges for shipping, that could change the ranking, so it's worth asking each of them.
You
Hmm, if we run out, B's 5 days is too slow. Can you give delivery time more weight than price?
Claude's reply
With delivery weighted above price, the ranking changes.
| Rank | Supplier | Delivery | Cost at 4 kg/month | Extra vs cheapest (B) |
|---|---|---|---|---|
| 1 | A | 2 days | £72 | +£12/month (£144/year) |
| 2 | C | Unconfirmed | £66 | +£6/month (£72/year) |
| 3 | B | 5 days | £60 | – |
- A is the safe pick. It's the only supplier with a confirmed fast delivery. The premium is £12/month, or about £1 per person per month, to avoid running out.
- C could beat A, but only if it confirms a delivery time of about 2-3 days. At 3 days or less it would save you £6/month and be nearly as safe. Without a confirmed date you can't count on that.
- B drops to last since you've said 5 days is too slow. Its price advantage doesn't matter if the office runs dry.
I'd go with A unless C confirms a delivery time close to A's. If you want to keep B's price, you could reorder when you have a week's supply left, but that only works if someone reliably does it. A removes that risk.
I'd also ask C for a delivery time and whether it guarantees it. A quoted time is worth less than a committed one.
You
Also, a colleague says Supplier A did a free trial for us last year and the beans tasted stale. Note that, but don't let it decide on its own.
Claude's reply
Noted. I'll treat it as a flag on A, not a disqualifier.
How much weight it deserves: It's a secondhand report of a single trial from a year ago. A trial batch may not match what you'd get on a regular order, and A may have changed roasters or processes since. But it does bear on A, because A's whole case rests on being the fast, reliable option, and stale beans would undercut that.
Where it leaves the ranking: The ranking hasn't changed, but A is less certain. A is still ahead on confirmed delivery (2 days), though the stale-beans report means it should earn its place rather than win by default. C moves from "maybe" to a more serious contender if it confirms a fast delivery time.
Ways to settle it:
- Ask A directly about the roast date. Do they roast to order, and will they guarantee beans shipped within a few days of roasting? Stale beans often come from old stock.
- Ask your colleague what "stale" meant. Flat taste with no aroma points to old beans. Bitterness or an off flavour could just be a roast style you didn't like.
- Order a small first batch from A and judge it yourself. At the 5 kg minimum that's £90, a low-risk test. Ask for the roast date on the bag.
- Request sample bags from C and A, and ideally B, so you're comparing current coffee, not last year's memory.
If A can't commit to a roast date, that tips things toward C. If C also confirms fast delivery, it could win outright at £6/month less.
You
Let's stop here for today. I'll get Supplier C's delivery time tomorrow.
Claude's reply
Sounds good. Here's where things stand for tomorrow.
Current position (4 kg/month, delivery weighted above price):
- A: £72/month, 2-day delivery. It's the safe pick for now, but a colleague reported stale beans from a free trial last year, so it needs a roast-date commitment.
- C: £66/month, delivery unconfirmed. It becomes the front-runner if it confirms about 2-3 days.
- B: £60/month, 5-day delivery. It's the cheapest, but you've ruled it out as too slow.
When you hear back from C:
- Get the delivery time, and ask whether it's guaranteed.
- Ask about the roast date and any shipping charges.
- If C is at 2-3 days, it likely wins at £6/month less than A. If it's slower, A is the pick, provided it can commit to fresh roasts.
If you'd like to continue, send me C's answer and I'll update the comparison.
The new chat's reply
Updated comparison with C at 3 days
At 4 kg/month, with delivery weighted above price:
| Rank | Supplier | Delivery | Monthly cost | Annual cost |
|---|---|---|---|---|
| 1 | C | 3 days | £66 | £792 |
| 2 | A | 2 days | £72 | £864 |
| 3 | B | 5 days | £60 | £720 |
Recommendation: C
- Delivery: A is one day faster, but 3 days is within the range I said would make C competitive. If you reorder with a week's supply left, a one-day difference is unlikely to matter.
- Price: C is £6/month (£72/year) cheaper than A.
- Freshness: A has the stale-beans report from last year's trial. It is secondhand and unconfirmed, but it is a small mark against A. C has no such flag.
- B: It is still out because of the 5-day delivery, even though it is cheapest.
A would only win if the extra day matters a lot to you, or if C can't back up its delivery claim.
Before you commit to C
- Is the 3-day delivery guaranteed, or just typical? A committed time is worth more than a quoted one.
- Roast date: Do they roast to order, and what is the typical gap between roast and dispatch?
- Shipping charges: I've assumed none. Any charge would narrow the £6/month gap with A.
- Taste: Ask for a sample bag, or place a small first order at C's 3 kg minimum (£49.50). That also tests the delivery time.
If C can't commit to a delivery time or roast date, fall back to A, provided A will commit to fresh roasts.
End of Claude's reply.
The new chat got only the handover: 440 words.
Show the handover it was given
Claude wrote this at the end of the original chat, in reply to the request from this post:
Write a handover for my next chat. Don't paste our chat; write only what a new chat needs. Start with a line on what this is, then these headings: Where things are, Current state, Rules, Next steps (in order).
Claude's handover
Handover: Choosing a coffee supplier for a 12-person UK office by comparing three quotes. The comparison was done in chat only, and nothing was saved or ordered.
Where things are
- Nothing has been written to a file, sent, or ordered. All the work so far is in the previous conversation, and this handover is the only record.
- The decision is waiting on one piece of information: Supplier C's delivery time. The user will get it from C.
Current state
Quotes:
| Price/kg | Delivery | Cost at 4 kg/month | Annual (48 kg) | |
|---|---|---|---|---|
| A | £18.00 | 2 days | £72 | £864 |
| B | £15.00 | 5 days | £60 | £720 |
| C | £16.50 | Unconfirmed | £66 | £792 |
- Usage is about 4 kg a month, which the user confirmed.
- Minimum orders (A 5 kg, B 10 kg, C 3 kg) are ignored, because the user has plenty of storage.
- Costs are for coffee only. Shipping charges are unknown for all three.
- B is out because the user says 5 days is too slow if they run out.
- A is the current default pick because it has the only confirmed fast delivery. The extra cost over B is £12/month.
- C beats A on price by £6/month, and becomes the likely winner if it confirms about 2-3 days.
- A colleague says A's free trial last year produced stale beans. This is secondhand, from one trial a year ago. It counts as a flag on A, not a deciding factor.
Rules
- Weight delivery time above price.
- Don't let the stale-beans report decide on its own.
- Don't assume C's delivery time. Treat it as unknown until the user supplies it.
- Don't factor in minimum order sizes.
- Keep answers short, with figures in a table where useful, and give a recommendation rather than a survey of options.
Next steps (in order)
- Get C's delivery time from the user, and whether it is guaranteed.
- Ask whether C or A charge for shipping, and update the monthly costs if so.
- Re-rank. If C is at 2-3 days, recommend C. If it's slower, recommend A, provided A can commit to fresh roasts.
- Suggest asking A and C for roast-date commitments, and for sample bags from each so they can taste current coffee.
- If the user wants more detail on the stale-beans report, ask what "stale" meant: flat and aromaless points to old beans, while bitter or off-tasting may be a roast-style issue.
The new chat's reply
Recommendation: go with C. At 3 days it falls within the 2-3 day range we set, and it's the cheaper of the two.
| Price/kg | Delivery | Monthly (4 kg) | Annual | |
|---|---|---|---|---|
| A | £18.00 | 2 days | £72 | £864 |
| C | £16.50 | 3 days | £66 | £792 |
| £15.00 | 5 days | £60 | £720 |
- C saves £6/month (£72/year) against A, and is one day slower.
- A also carries the stale-beans flag from the colleague's trial. It isn't decisive, but it adds to C's case.
- Delivery time still outweighs price here. If one day matters to you, A is the fallback, but I wouldn't pay £72/year for it.
Before you commit:
- Ask C whether the 3 days is guaranteed or just typical.
- Ask C (and A, if you keep it as a backup) whether shipping is charged. That could shift the figures above.
- Ask C for a roast-date commitment and a sample bag, so you can taste current coffee before ordering.
Can you check the shipping charges and the guarantee with C?
End of Claude's reply.
With a format like this between each step, you can split a big job into steps that each do one thing, such as research in one chat, a first draft in the next and checking in a third, with each chat starting from the last one's handover. I think of an assembly line for cars, where everyone works on the same conveyor belt and each person has a specific job as the car moves along it. Here each chat does one job, and the handover passes the work to the next chat. If one step goes wrong, you redo just that step from the last handover, and the steps before it stay as they are.
Get it to do an export out in case Claude falls down tomorrow.
My main reason for all this is not losing work when things are offline. Chat apps keep changing, so I don't keep work only inside them, and I ask for an export out.
Between colleagues, a common way to share AI work is to summarise a chat and paste the summary into someone else's chat, which means copying and pasting back and forth. A simpler way is for each person to end the day by asking for a short daily log and adding it to the team's shared project, so every chat in that project can read everyone's logs without anyone pasting anything.
Write a short Markdown note of what I did today and what's next, for my team's daily log.
Upload the log it writes to the shared project, or paste it in as text.
If you haven't shared a project before, Shared project instructions: one rulebook for your team's AI shows how in Claude, ChatGPT and Gemini. You can then open a chat in the shared project and ask its AI to scan the project's files for updates from the team and bring you up to speed.
Members can get the Handover kit from the Members Library. It has the handover request, the format file, steps for keeping the format in Claude, ChatGPT and Gemini, prompts for a team daily log and a handover skill for Claude Code.
At the end of your next chat, paste the handover request above. If you work in a team, add today's daily log to the shared project too.