Layla Hassan was eighteen months into a sociology PhD when she realized her thesis was going to miss its submission deadline — not because of a lack of ideas, but because of a planning failure.
She had extensive notes. She had read deeply in her field. She had a supervisor who gave good feedback. What she did not have was a functional system for turning three years of accumulated material into a coherent 80,000-word document within a fixed timeline.
This is a composite case study drawn from patterns common among PhD students using AI for planning. The methods are specific and replicable.
The Problem: Unstructured Abundance
Layla’s situation is common among graduate students who have done the reading. The notes are there. The ideas are there. The analytical muscle has been developing for years. But the path from “I have a lot of material” to “I have a complete thesis” is not obvious, and the timeline of a PhD makes it easy to defer the hard work of structuring.
When Layla started using AI for planning — not writing — the first conversation was about diagnosis, not solutions.
I'm a PhD candidate in sociology, 18 months in. My thesis examines
how informal care networks among immigrant communities in [city]
substitute for formal social services. I have:
- 14 months of ethnographic fieldnotes
- 40+ interviews (conducted, some not yet coded)
- A literature review draft that is scattered and too long
- A draft introduction that I rewrote three times and still hate
- 14 months until submission
My supervisor says my ideas are strong but my structure is weak.
I'm not asking you to write anything. I need help thinking through
my planning problems:
1. How do I turn this material into a structured thesis plan?
2. What should I do in what order over the next 14 months?
3. What are the most common planning mistakes PhD students make at
this stage?The AI’s response to question 3 was the most useful entry point: it named the common failure as “writing without a map” — beginning to draft chapters before the overall architecture of the argument is settled. Layla recognized herself immediately.
Step 1: Building the Argument Architecture
Before any writing, Layla needed a clear answer to a prior question: what is the central claim of this thesis, and what does the reader need to believe at each stage to accept it?
She used AI as a thinking partner to work this out — but crucially, she did the intellectual work. AI asked questions; she answered them.
I'm going to try to articulate my central argument in one sentence.
Then I want you to ask me the hardest questions an examiner might
ask: what assumptions I'm making, where the evidence might not
hold, and what alternative explanations I haven't addressed.
My central argument is: [Layla's draft argument]
Start with the question you think is most likely to expose a
weakness in this framing.Over three sessions, this conversation produced a sharpened central argument and a list of the six most significant challenges Layla needed to address in the thesis. That list became the backbone of her chapter structure.
Nothing AI produced here was submitted. The clarified argument and the chapter structure were Layla’s — developed through her own answers to AI’s questions.
Step 2: The 14-Month Working Backwards Plan
With an argument structure in place, Layla needed a monthly and weekly plan.
My thesis has the following structure:
- Introduction (chapter 1)
- Literature review (chapter 2, currently messy draft)
- Methodology (chapter 3, mostly done)
- Four empirical chapters (4–7, largely in notes form)
- Conclusion (chapter 8)
- Submission deadline: [date, 14 months away]
I have 3–4 writing hours per day, 5 days a week. I also need to
code 15 remaining interviews (I estimate 2 hours each).
Please build a month-by-month plan that:
1. Gets interviews coded before deep writing begins
2. Attacks the weakest chapter (literature review) while I have
the most energy, not last
3. Builds in one revision cycle for each chapter before final assembly
4. Leaves the last two months for full-document review and
supervisor feedback turnaroundThe AI produced a phased plan with monthly milestones and a daily word-count target that was achievable rather than aspirational. Layla adjusted two months based on conference commitments and field work, but the skeleton held.
She tracked her actual time against plan using Beyond Time, which let her see which chapters were taking longer than estimated and adjust the downstream schedule before the overrun became a crisis.
Step 3: The Daily Writing Session Design
A 14-month plan is only useful if the daily sessions are productive. Layla had a second planning problem at the micro level: she was frequently sitting down to write and spending the first 45 minutes deciding what to write.
She built a session design ritual using AI:
I'm starting a 3-hour writing session. I'm working on chapter 5
(empirical chapter on [topic]).
My outline for this chapter has the following sections:
[list sections with rough word targets]
I've completed sections 1 and 2. Section 3 is the hardest —
it requires synthesizing interview data with theoretical claims
I'm not confident about.
Design my 3 hours:
1. What should I tackle first to use my highest-focus time well?
2. How should I handle the section 3 difficulty without getting stuck?
3. What is a reasonable word count to have produced by the end of
this session?This two-minute conversation replaced the 45-minute drift. Layla started each session with a specific sequence rather than an open-ended intention.
Step 4: Using AI as an Argument Stress-Tester
The Feynman principle applies to academic writing as much as to exam preparation. After drafting each chapter section, Layla used AI to challenge her argument — not to revise her writing, but to identify logical gaps she needed to address herself.
Here is the argument I've made in section 3 of chapter 5.
I am not asking you to rewrite it.
[Pastes her own text]
Please:
1. Identify the two places where my reasoning makes the biggest
inferential leap
2. Ask me the questions a skeptical examiner would ask about
my evidence for each of those claims
3. Do not suggest how to fix the problems — just identify them
clearly so I can decide how to address them in my own revisionThis is using AI to improve your own work without substituting for it. The distinction: AI is finding the gaps; Layla is filling them.
What Layla Submitted (And What She Didn’t)
At submission, every word in Layla’s thesis was hers. AI had:
- Helped her clarify her argument through questions
- Built her 14-month working schedule
- Designed her daily writing sessions
- Identified logical weaknesses for her to address
AI had not:
- Written any sentences she submitted
- Summarized sources she then cited
- Produced her literature review or any analysis
- Contributed any intellectual content to the thesis
Her examiners found the structure unusually clear and the argument well-defended. One noted that the thesis “had the feel of a writer who knew exactly what she was arguing throughout.”
That clarity came from the planning work — from having a map before starting to write. AI provided the planning infrastructure. Layla provided the scholarship.
The Replicable Elements
Any graduate student can use the same pattern:
- Use AI to stress-test and sharpen your central argument before writing begins
- Build a backwards-planned schedule from submission date with monthly milestones
- Design each writing session with a specific sequence and target before you start
- Use AI to identify argument weaknesses in your own drafts — not to fix them
The line is consistent throughout: AI helps you plan and challenges your thinking. You do the intellectual work. The thesis is yours.
Build your 14-month (or 14-week) backwards plan this week. Start from your deadline and work back to tomorrow.
Tags: grad student AI planning, PhD thesis planning with AI, graduate student productivity, AI for academic writing
Frequently Asked Questions
-
Can PhD students legitimately use AI in their research process?
Yes — at the planning, structuring, and thinking stages. Using AI to build a chapter outline, stress-test an argument, or build a research reading schedule is legitimate. Using AI to paraphrase sources you then cite, generate your literature review, or write your analysis is not. Most graduate programs have specific policies, and the standards are higher than undergraduate courses because the work is supposed to represent original scholarly contribution.
-
How did the student in this case study use AI without crossing academic integrity lines?
Every AI interaction was at the planning and scaffolding layer. AI helped her build schedules, stress-test her argument structure, and generate self-test questions. AI never produced content she submitted — not a sentence, not a paraphrase. The thesis is entirely her writing, argument, and analysis. AI was the planning infrastructure, not the work itself.