thinkhard.SI

Prompt repository Operations

Draft a service level agreement between two internal teams

One team depends on another for requests such as IT tickets, purchasing or facilities work, and you need response times, priorities and escalation in writing.

Registry no.
OPS-031
Collection
Operations
Laid out for
Any chat model
Score
10 out of 10
Blanks to fill
8
Added

OperationsAny chat

The prompt

Score 10 of 10, Excellent

Use it when One team depends on another for requests such as IT tickets, purchasing or facilities work, and you need response times, priorities and escalation in writing.

  • {{provider_team}}
  • {{customer_team}}
  • {{service}}
  • {{agreed_points}}
  • {{current_performance}}
  • {{service_hours}}
  • {{pain_points}}
  • {{word_limit}}
Full prompt and why it works
Prompt
Draft an internal service level agreement between {{provider_team}}, who provide the service, and {{customer_team}}, who use it, for both team leads to review and sign, so everyone knows what to expect from {{service}} and what to do when it goes wrong.

What has been agreed or proposed so far:
"""
{{agreed_points}}
"""

Current performance, if measured (volumes, response and completion times):
"""
{{current_performance}}
"""

Hours the service runs: {{service_hours}}
Known pain points on both sides: {{pain_points}}

Use these headings:
1. Service covered: what is included and what is not.
2. How to make a request: the channel and the information every request must include.
3. Priorities: a table with the columns Priority, Definition with an example, Response time, Target completion time.
4. Responsibilities of each team, including what the customer team must provide for the targets to apply.
5. Measures and reporting: what is measured, how, how often and who sees it.
6. Escalation: the steps and the roles involved when a target is missed.
7. Review: when the agreement is reviewed and how changes are agreed.

Keep it under {{word_limit}} words, in plain language both teams can follow without training. Every target needs a number and a unit, for example "Priority 2: response within 4 working hours, fixed within 2 working days". This example shows the form only.

Use only targets from the agreed points. Where a section needs a target or rule nobody has agreed, write [to agree] and, if current performance is given, add the range it suggests, rather than choosing one. If a proposed target is tighter than current performance shows the team achieves, flag it in a short note at the end, because a target missed from the first week undermines the whole agreement. Do not add charges or penalties between teams unless the agreed points include them.

Variables

{{provider_team}}
The team providing the service, for example "IT support" or "the facilities team".
{{customer_team}}
The team using it, for example "the regional sales offices".
{{service}}
The service covered, for example "laptop and access requests".
{{agreed_points}}
What both sides have agreed or proposed: targets, priorities, channels, exclusions.
{{current_performance}}
Current volumes and times if you measure them. Write "not measured" if you do not.
{{service_hours}}
When the service runs, for example "8am to 6pm, Monday to Friday, excluding holidays".
{{pain_points}}
What goes wrong today from each side, for example "requests arrive without approvals; no updates on progress".
{{word_limit}}
The most words you want, for example 900.

Why it works

  • What good looks like. Every target needs a number and a unit, which is what lets either team tell whether the agreement is being kept.
  • If unsure. Unagreed targets are left as [to agree] with the range current performance suggests, so the draft does not settle what the teams have not.
  • What to avoid. No charges or penalties unless both sides agreed them, which keeps the draft from creating a fight the leads did not ask for.

How to use it

  1. Press Fill in variables and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/internal-service-level-agreement

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "internal-service-level-agreement"}, or just ask for it by name.