Openwork and Fieldfisher rollouts: test the changed instruction
Openwork and Fieldfisher announce different stages of technology rollout. Test whether a changed client instruction stays traceable from receipt to action.
The brief
Openwork has announced an adviser technology suite, while Fieldfisher is rolling out Legora across its international offices. Openwork’s release puts early availability in the first half of 2027; Legora’s announcement follows a six-week Fieldfisher pilot. These are useful signs of investment in connected workflows, but they are different stages of delivery. Our proposed acceptance test follows one changed client instruction from receipt to review and execution. Can the team retrieve the evidence behind the action after information moves between systems? Test a late correction and a withdrawn instruction as well as the ordinary path. A rollout count or a promised time saving should not substitute for an observed result on that test.
The full record
On 3 November, Openwork announced a suite built on Plannr and Seccl, with wider rollout planned from the second half of 2027. Its claim of up to an hour saved per case is an announced ambition, not an independently measured outcome reported here. Openwork announcement.
Legora said on 5 November that Fieldfisher was extending its platform to around 1,400 lawyers and business-services staff after a 137-user pilot. The release describes training and shared workflows under the firm’s AI governance. It does not publish a comparative error rate or a quantified improvement in client-instruction handling. Legora announcement.
Our analysis: follow a changed instruction
Use a fictional matter in which the client first approves one destination for a document, then changes it before dispatch. Let the first instruction travel through the planned workflow. Introduce the correction at the point where another team would ordinarily act. Ask which instruction the system presents as current, who must review the change and what stops the earlier version from being used.
The test should include the eventual reviewer. Give them the completed file without a verbal explanation from the people who handled it. Can they reconstruct what the client asked, what changed, who approved the action and which evidence supported it? If they need a participant’s memory, the record is incomplete for that review purpose.
Keep the measurement tied to the task
Record elapsed time, manual corrections, missing evidence and any action taken on an outdated instruction. Compare the same scenario before and after a change, with the same scope and clear rules for counting work. Report preparation and training time separately from processing time. Avoid turning a faster demonstration into a claim about all matters.
Use failures to choose the next change: a missing instruction version calls for a version link; unclear ownership calls for a named reviewer; a stale destination calls for a stop and reassessment. Buying more integrations alone does not establish that any of those problems is resolved.
This proposed test is our analysis. We have not tested either platform or audited either organisation. The announcements justify asking precise procurement questions; they do not answer those questions on the organisations’ behalf.
We read the company and vendor announcements. These establish announced plans and deployment scope, not independent performance. The workflow exercise is an original proposal and has not been conducted. No causal claim about productivity or fraud prevention is made.