A contract goes out to a client. They send it back with changes. Someone on your team has to read both copies, find every difference, weigh the impact, and decide what to push back on. Next week the same task lands on someone else and the output looks different, because the process lives in the reviewer's head.
I built a SharePoint AI Skill that runs the gaps analysis end to end. The Skill loads the base contract automatically. The user selects the client's returned copy. The output is a traffic-light table of every change, priority-ranked by business impact, with an executive summary and a recommended action plan. Anyone on the site with View permission can run it.
This is the build walkthrough. You can apply the same pattern to any review task your team runs by hand every week.
If you have not read what SharePoint AI Skills are, start there. This piece assumes you know the basics and want to see a real one shipped.
The problem this Skill solves
Contract review by hand has three failure modes. It is slow. It is inconsistent across people. And it usually drops the prioritisation step entirely when the reviewer is rushed.
Every contract review done by hand goes the same way. Read the base. Read the returned copy. Spot the changes. Try to remember which ones matter most. Write up a summary. Decide which clauses to push back on. Repeat next week with someone else, who does it slightly differently.
Skills replace the human-memory layer with a saved instruction. Once the Skill is written, the same logic runs every time, regardless of who invokes it.
What the Skill does in plain English
The Skill loads the base contract automatically from the site's documents library. The user selects the client's returned copy. The Skill performs a clause-by-clause comparison and presents the result in a single table.
For each change it captures the original clause, the client's modified version, the business impact, and a priority ranking for negotiation. The table uses traffic-light colours so the highest-impact changes stand out at a glance.
Below the table the Skill writes an executive summary of the overall risk, a list of top business and legal impacts, and a priority action plan. All in one run, all from one prompt.
Setup: enable the Agent Assets feature
Skills live in a system library called Agent Assets. The library is created when you activate a site collection feature with the same name.
To turn it on:
- Open Site information on the site.
- Click View all site settings.
- Open Site collection features.
- Find Agent Assets at the top of the list and activate it.
The library appears immediately. Inside it you will find two folders: Skills and Plans. The contract Skill lives in the Skills folder. For the full mechanics see Microsoft Learn on extending AI in SharePoint with Skills.
Writing the prompt that builds the Skill
I do not write Skills by hand. I describe what I want in chat and let AI in SharePoint draft the Markdown. The prompt I used for this one.
I want to create a Skill that performs a gaps analysis on contracts.
We have a base contract stored in the documents library called
"Base Contract.docx". I want this contract loaded automatically
every time the Skill runs.
Users send this contract to clients. Clients return it with changes,
amendments, or deletions. We need to find every change and assess
the impact on our business.
Output:
- A traffic-light table of every change.
- An impact analysis covering legal, business, and reputational risk.
- A priority ranking for negotiation, highest impact first.
The agent drafts the Skill, asks two clarifying questions (does the user select the client file, and do I want a recommended negotiation response per item), and writes the Markdown to /Agent Assets/Skills/Contract Gaps Analysis/SKILL.md once I confirm.
Running the Skill on a real contract
After saving, anyone with View permission can run the Skill. They open the documents library, select the client's returned contract, and prompt AI in SharePoint with something as simple as "run a contract analysis".
The Skill loads (the chat shows a Skill indicator card to confirm). It fetches the base contract from the library. It runs the clause-by-clause comparison. It produces the output.
On my first real test, the Skill flagged 24 distinct changes. The top of the table held the highest-impact ones: payment terms, liability, indemnity. The bottom held minor wording tweaks. Everything in between was ordered by business impact, which is what I asked for in the prompt.
The output you should expect
The output is one structured response in chat. Top to bottom:
- An executive summary, four or five lines, on the overall risk profile of the changes.
- The traffic-light gaps analysis table. Columns: original clause, client change, business impact, priority. Rows ordered by priority.
- An impact analysis summary covering legal, business, and reputational risk.
- A priority action plan listing every change with a recommended response.
The same structure runs every time. If a junior team member runs the Skill while I am away, the output lands the same way it would if I had done it myself.
Why this matters more than it looks
The Skill is not the productivity gain. The standardisation is.
Before the Skill, the gaps analysis varied by reviewer. One person might miss the priority ordering. Another might write a tight summary but skip the impact analysis. The output depended on the reviewer's mood, time pressure, and memory of the playbook.
With the Skill, the output is the same shape every time, by every reviewer. The expert writes the SOP once. The team inherits it. That is the quiet shift Skills make in any review workflow that runs every week.
Where to take this next
Pick the review task your team runs every week by hand. Document review, contract review, RFP scoring, proposal QA. Write the prompt the way I wrote it above. Test on three real files. Save it.
You will have a Skill shipped before lunch. The second one is faster than the first.