From request to agreement: reducing manual work

How the process worked, and what slowed it down

Swissnex in China connects Switzerland and China in education, research and innovation. Universities, startups and companies approach the team for introductions, exchanges and tailored programmes.

A project request arrived by email. A team member was assigned according to expertise and capacity, then arranged a meeting to understand the request. The proposal was drafted manually, reviewed by management and sent to the requesting organisation. Once both sides agreed, the proposal became the basis for a manually drafted contract, followed by review and signature.

Alongside this, the team entered information in Zoho CRM, which records organisations, contacts and project history, and in management tracking sheets. Changes had to be carried between emails, documents and both systems.

The problem was repeated work throughout the process: typing the same information, assembling it again for each document, chasing missing details and keeping versions consistent. This reduced the time available for delivering projects and maintaining relationships.

The proposed improvement connects the whole process. Information captured from the request and meeting supplies the proposal, the accepted proposal supplies the contract, and the CRM supplies management's planning view. AI handles varied text and drafting; templates and rules handle predictable tasks; people confirm requirements and approve commitments.

The corporate scouting example below is fictional and illustrates the integrated design. Measured results from the pilot are reported at the end.

1. Turn the email into a prepared project record

Requests can still reach any team member. Forwarding a project enquiry to a dedicated address starts the workflow.

AI reads the email and prepares a draft CRM entry: the organisation, contacts, requested activity, timing and missing information. The entry is linked to the existing organisation where possible, preserving previous interactions.

The workflow assigns a team member using agreed expertise and capacity criteria. The Head of Programs receives a notification and can reassign the request without making every routine allocation wait for approval. Projects involving several colleagues have one lead and clearly identified contributors. The assigned team member checks the extracted information and confirms the correct CRM record.

Consider a company requesting a three-day innovation scouting visit to China. The record captures that request immediately, while dates, budget and the programme's objective remain questions for the meeting. The team checks a prepared record instead of creating it from scratch and copying it elsewhere.

2. Arrange the meeting with simple automation

Once the owner and meeting participants are settled, a fixed email template offers times drawn from their calendars. This uses Gmail and Google Calendar alongside the CRM.

AI adds little to a standard scheduling message. Calendar rules can offer available slots, place temporary holds and remove unused holds once a time is chosen. Unanswered offers expire so calendars do not remain blocked.

This removes routine email preparation and helps prevent double booking. Meanwhile, the owner prepares for the meeting using the enquiry and relationship history already collected in the project record.

3. Use the meeting to prepare the proposal

The meeting establishes what the organisation needs to achieve. That remains a human task: asking the right questions, applying local knowledge and deciding what can realistically be delivered.

In the scouting example, the discussion reveals that the company wants technical partners for automating quality inspection. That gives the programme a purpose: targeted meetings with relevant companies and research groups, with clear follow-up, rather than a broad visit.

An approved AI note-taking tool turns the discussion into a draft brief containing the objective, activities, deliverables, responsibilities, timing, budget and open questions. Written notes can provide the input where recording is unsuitable. The owner checks the brief and confirms the decisions.

AI then drafts the proposal from the checked brief, using reusable instructions and an approved template for its content and layout. The same brief updates the CRM. Missing information is flagged; dates, prices and commitments must come from confirmed inputs.

The owner checks the draft's accuracy and feasibility, resolves open points and submits it to management. Only the approved version is sent to the partner.

One checked brief would serve both project documentation and proposal preparation, reducing assembly and formatting work. Clarifying the organisation's goal before drafting should also reduce avoidable proposal revisions.

4. Carry agreed changes into the contract

Email exchanges stay linked to the project. AI can identify requested changes and prepare updates, which the owner confirms before they alter the agreed requirements.

If the company asks to shorten the visit from three days to two, the owner checks what this means for the programme, deliverables and price. The revised proposal follows the required approval, and the version accepted by the partner becomes the basis for the contract.

AI drafts the contract from that accepted proposal, using the standard structure and lawyer-approved wording. It recommends clauses suited to the project. A missing clause or substantive change to approved wording triggers legal review.

The reviewer sees the accepted commitments, the selected clauses and highlighted edits. Each clause's applicability must be confirmed, corrected or referred for review. This helps prevent a polished AI draft from being accepted without scrutiny. Management approval follows, then the agreement is sent to the partner for approval and signature.

The contract uses the final two-day scope and agreed terms without rebuilding them from the email chain. Drafting is reduced, while decisions about commitments and legal wording remain with the appropriate reviewers.

5. Replace the duplicate tracking sheets

The same CRM record can show the project owner, stage, next action, deadline, resource needs and pending approvals. Management gets a planning dashboard from those fields.

Before retiring the sheets, identify the information management actually uses and add anything missing to the CRM view. Test that view with the team, agree when it becomes the normal tracking method and archive the old sheets.

For the scouting project, recording that the proposal awaits approval also updates management's queue. After approval, the next action becomes sending it. The team records the change once.

Introduce the change, then check the result

Start with one recurring project type. Prepare its CRM fields, proposal template and approved clause library, then test the workflow on representative cases.

Involve the people doing the work: identify what they currently repeat, check what the new process removes and make the remaining review responsibilities clear. A pilot should include incomplete requests, returning partners and scope changes. If an automated step fails, the team should be able to continue manually in the same project record.

Measure preparation time including checking and correction, proposal revision rounds and the full time from request to signed agreement. Compare simple and complex projects separately. Faster drafting will not remove every wait for management or partner approval, so the full process must be measured.

Pilot results

For comparable straightforward projects used in the pilot, the redesigned process reduced preparation effort from 8 to 4 hours per project (50%), including checking and correction, and shortened request-to-signature time from 15 to 12 working days (20%).

Clearer proposals and more reliable scheduling should give partners a more responsive service, with less back and forth and fewer avoidable rescheduling problems. Less administration left more time for project delivery and the human relationships that generate future projects.