In adopting legal tech (and in particular of AI), several areas of tension surface. Just like hiring extra legal counsel frustrates adoption of legal tech, serving your internal clients in the best way possible may well be counterproductive for AI or legal-tech adoption. Where hiring (instead of automating) conceals a ‘people-readiness’ gap, a customer-first approach likely conceals ‘process-readiness’ gaps.
Broken processes
Supporting your internal stakeholders in the best way possible, may well imply ‘broken processes’. Over the past months, I have seen Harvey and Legora proudly sharing phenomenal numbers of ‘workflows’ they had configured or implemented. But is that so praiseworthy, a high number? In times of AI, for a legal AI provider, shouldn’t the number be as low as possible?
What is a broken process?
Imagine how a simple contract may originate and find its way to signing:
- Asking you by forwarded e-mail (with incomplete info).
- Asking you via Teams (requiring you to enquire).
- Passing by your office to ask (giving lots of insights but no hard details).
- Asking their team member (off the radar process).
- In your absence, asking your colleague (who will both help and hand over to you).
- Using their previously used template (off the radar process).
- Asking you to handle it.
- Expecting to handle it entirely themselves.
- Starting with the counterparty’s template.
- Starting when the deal is done (at least, momentum is about to lapse).
Isn’t the above list absurd? A broken process is where an end-goal (a signed, unequivocally clear contract that helps everyone involved) is required but where involvement of people, or the collection and management of contract (meta)data follow somewhat random patterns to get there.
Checks and balances
Ten ways to initiate, process and sign a contract. Imagine that your organisation has multiple departments and BUs, each processing their contracts slightly differently:
- Varying signing authorisations.
- Varying approval processes (reflecting different leadership styles).
- No unequivocal ‘ownership’ for such contract (as it is part of other workflows).
- No uniform insights at management levels.
And as part of the signing process, where (unfortunately) many precontractual aspects are both revealed and being settled, how does registration and management of a signed contract happen?
- Registration is done at the secretariat of Legal (basics only: expiry notification).
- Contract management is a spreadsheet or Outlook calendar- or task-item.
- Contract management is a CRM-entry (but without the contract’s documentation or without full-text search capabilities).
- Non-standard deal-particularities are the responsibility of (the manager in) the business (including preventing the potential revenue leakage).
All deviations may be for good reasons, but no such process follows a predefined workflow. Automating such process is appealing for all stakeholders involved, but who will fix the process? Or will everyone be allowed their own process?
Processes and RevOps
Each process is at best designed in the best interest of the department concerned (sales, procurement, operations, customer delivery), but not necessarily for ‘Rev Ops’. RevOps considers an organisation’s orientation on the core processes spanning all of marketing, communication, sales, delivery and after-sales support. Those working on marketing and communications should know what happens after sales, and customer account managers must not be unaware of the sales managers’ promises or activities on social media.
RevOps requires that data and people are aligned on what matters to the organisation as a whole. RevOps is a proper (orgainsational) prerequisite for an effective implementation of agentic AI, like solid Legal Ops is a proper prerequisite for legal-tech or AI adoption.
Processes and agentic AI
When the organisation prioritises agentic AI, it will automate processes as much as possible. After all, an automated process can be performed by an AI agent.
Now, a legal AI provider that applauds their number of implemented workflows, is suspicious of having overlooked the importance of aligning processes (rather than extrapolating them), of reducing the number of workflows (rather than revealing how little uniformity there is).
While it is almost impossible to optimise a process without automating it (to experience the urgency, and to learn what the optimal process is), making any process (AI) agentic or implementing AI agents, is a no-go area for organisations that are simply ‘not ready’ for tech, that have too many maturity gaps in people, data or processes.
Any idea how to implement AI and move towards agentic AI?
