Skip to content
Curious Context Labs

case study · Operations · Company assistant

Answers from the right procedure, on the shop floor, in seconds

How an assistant built on your own SOPs can answer frontline questions with a link to the source, and say when it does not know.

business
Manufacturing
build
Company assistant
timeline
4 weeks
read
3 min

01 the business

Precision engineering plant

180 staff across three shifts

Procedures sit in binders and shared folders that nobody opens mid-shift. New starters ask whoever is nearby, and the answer depends on who that is. Supervisors keep being pulled away to repeat the same instructions.

02 before

The problem

Procedures live in binders and shared folders that nobody opens mid-shift. New starters ask whoever is nearby, and the answer depends on who that is. Supervisors get pulled away from their own work to repeat the same instructions, and steps get skipped.

  • Answers vary by who is asked.
  • Supervisors lose time to repeated questions.
  • Missing or unclear procedures are never spotted.

03 the build

What the AI does

  1. Step 1: Reads your SOPs, checklists, equipment manuals and safety sheets from SharePoint, Google Drive or Notion.
  2. Step 2: Answers questions in Microsoft Teams, Slack or a simple mobile web page, in short numbered steps.
  3. Step 3: Links to the exact procedure and page each answer came from, so staff can check the source.
  4. Step 4: Answers in the staff member's preferred language when your team speaks more than one.
  5. Step 5: Says when it does not know and points the person to their supervisor, instead of guessing on safety-critical tasks.
  6. Step 6: Logs the questions it could not answer so managers can see which procedures are missing or unclear.

paused: waiting for a person to approve

How it was put together

  • Loaded approved SOPs, checklists, manuals and safety sheets from SharePoint.
  • Answered in Teams and on a mobile web page in short numbered steps.
  • Linked every answer to the exact document and page.
  • Logged unanswered questions for managers to review.

04 human in the loop

Where a person stays in charge

Managers decide which documents the assistant can use and approve updates to them. Safety and clinical procedures are answered word for word from the approved document, never paraphrased.

  • Only approved documents are used; managers control the list.
  • Safety-critical procedures are quoted word for word, never paraphrased.
  • When unsure, it says so and points to the supervisor.

05 rollout

From kickoff to live in 4 weeks

  1. Weeks 1–2Gather and clean the documents; retire outdated versions.
  2. Week 3Pilot with one shift and its supervisor.
  3. Week 4Roll out to all shifts and start the monthly unanswered-questions review.

06 after

What changes

Staff get the right answer in the moment, and the answer is the same whoever asks. Supervisors spend less time repeating instructions, and gaps in your procedures become visible.

worked example · your numbers will differ

Assume 180 staff each asking a supervisor two questions a week, at 5 minutes per interruption.

180 × 2 × 5 minutes = 1,800 minutes, or 30 hours a week of supervisor time.

What to measure

  • Questions answered with a source link.
  • Questions it could not answer, by topic.
  • Time to competence for new starters.
  • Supervisor interruptions per shift.

07 evidence

What the research says

  1. Across 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14% on average and by 34% for novice and lower-skilled agents.

    NBER, Generative AI at Work (Brynjolfsson, Li and Raymond) (opens in a new tab) · 2023

  2. Knowledge workers spend roughly a fifth of their working week searching for and gathering information.

    McKinsey Global Institute, Capturing business value with social technologies (opens in a new tab) · 2012

  3. 62% of people said they spend too much time searching for information, 68% lack enough uninterrupted focus time, and 55% said next steps at the end of meetings are unclear.

    Microsoft, 2023 Work Trend Index: Will AI fix work? (opens in a new tab) · 2023

08 lessons

Mistakes to avoid

  • Loading out-of-date documents. Clean first.
  • Paraphrasing safety steps. Quote them exactly.
  • Ignoring the gaps log. It tells you which SOPs to fix.

09 questions

Common questions

Does research support this?
A study of 5,179 support agents found AI help raised productivity most for newer, less experienced staff, by 34%. New starters benefit most here too.
Can it answer in other languages?
Yes, while still linking to the approved source document.

keep reading

All case studies ->

Tell us how your business runs. We'll tell you where AI fits.

A free 30-minute call. You talk, we ask questions, and you leave with a few concrete ideas — or an honest answer that AI isn't the right fix yet.

Working with teams around the world.