Accountability and Your 90-Day Plan
Module 6 · Communication and Accountability

Accountability and Your 90-Day Plan

Own your AI-assisted work through documentation, transparency, and escalation; help your team adopt AI responsibly; and build a 90-day plan that applies all four fluency competencies.

Your Name Is on It

Every lesson in this course comes back to the same point: AI can draft, but it can’t be accountable. Your signature on a batch record, your name on a paper, your approval on a safety case, and your sign-off in MLR all mean you stand behind the content, however it was produced.

Accountability in practice comes down to four habits.

1. Document where it matters

You don’t need to log every grammar fix. You do need a record when AI materially shaped something that’s GxP, submitted externally, published, or used for a significant decision. A practical minimum:

  • What tool and version were used.
  • What it was asked to do, and what inputs it received.
  • What a person checked, and how.
  • Who approved the final result.

Follow your SOPs. Where none exist yet, this list is a reasonable default.

2. Be transparent

  • With colleagues: say when a draft was AI-assisted, so reviewers know what to look for.
  • With journals and funders: follow their disclosure policies.
  • With regulators: follow your sponsor’s position, and engage early when AI supports a regulatory decision.
  • With patients and participants: be honest when AI shapes what they receive, in line with consent and applicable law.

3. Escalate problems

If you find an AI error that reached a GxP record, a submission, a publication, or a patient, treat it like any other quality event: report it through the normal channel, such as a deviation, CAPA, or safety report. Early reports of AI failures are how organizations learn where the controls are weak.

4. Keep learning

Tools, guidance, and regulations covered in this course will change. Several were drafts when this course was written. Make a habit of checking for updates in the areas you touch.

⚠️ Common Pitfall: Shadow AI Becomes the Real Process

Common mistake: a team quietly adopts an unapproved tool because it’s faster, the practice spreads, and eventually a deliverable depends on it, with no validation, no data agreement, and no record.

If a tool is genuinely useful, make the case to get it approved. Shadow use doesn’t just carry risk. It also stops your organization learning what people actually need.


Helping Your Team Adopt AI Well

If you lead a team or influence how it works:

  1. 1 Inventory

    What's already in use?

  2. 2 Classify

    Risk and GxP status

  3. 3 Approve

    Tools and data rules

  4. 4 Train

    Fluency, not just features

  5. 5 Measure

    Time saved and errors caught

  6. 6 Revisit

    Quarterly review

  • Start with low-risk, high-volume tasks where errors are easy to catch, to build skill and trust.
  • Share failures as well as successes. A team channel of “AI got this wrong” examples is one of the best training tools there is.
  • Measure honestly. Count time saved, and also time spent checking and errors caught. A tool that saves an hour and costs 50 minutes of review is a smaller win than it looks.
  • Bring quality and compliance in early. They are much easier partners at the design stage than after an audit finding.

Your 90-Day Plan

Delegation pick the right tasks Description build reusable prompts Discernment set verification habits Diligence document and share Days 1–30FoundationClean up tool inventoryRead policy and SOPsPick 2 low-risk tasksDays 31–60Build the habitWrite reusable promptsVerify every timeLog time and errorsDays 61–90Extend and shareAdd 1 moderate-risk taskShare wins and errorsStart 1 approval talkDay 0Day 30Day 60Day 90Day 90 review: keep, change, or stop

Days 1–30: Foundation

  • Revisit your AI inventory from Lesson 1.1. Remove or replace any unapproved tools.
  • Find your organization’s AI policy, approved tools, and the SOPs that apply to your role. Note the gaps.
  • Choose two low-risk, high-volume tasks to use AI for consistently.

Days 31–60: Build the habit

  • Write a reusable prompt for each of your two tasks using the structure from Lesson 3.1.
  • Define the verification step for each and actually do it every time.
  • Keep a simple log: time saved, checking time, and errors caught.

Days 61–90: Extend and share

  • Add one moderate-risk task with a documented review step, after checking with your quality or compliance partner if it touches GxP.
  • Share what you’ve learned, including errors, with your team.
  • Identify one process where AI could help but would need validation or approval, and start that conversation.
🧪 Try It: Write Your Plan

Draft your 90-day plan on one page:

  1. Your two starter tasks and the verification step for each.
  2. The one moderate-risk task you’ll add, and who you’ll consult.
  3. The policies and SOPs you need to read or request.
  4. How you’ll measure whether it’s working.
  5. A date, 90 days from now, to review the plan and decide what to keep, change, or stop.
🧠 Knowledge Check

1. You find that an AI-generated error made it into a signed GxP record. What should you do?

2. Which tasks are best for starting a team's AI adoption?

3. Why should time spent checking AI output be measured alongside time saved?

Course Complete

You now have a working framework for using AI across the life-science lifecycle: know the failure modes, delegate by risk, protect sensitive data, verify against primary sources, respect GxP and regulatory expectations, and stay accountable for everything you sign. The tools will keep changing. Those habits won’t.