Every enterprise SharePoint has a folder where good intentions go to die.
You know the one. It’s called something like Process Documentation 2023 or 2021, or 2019, depending on when the last sponsor lost interest. Inside: a few hundred PDFs nobody reads, a Visio file that won’t open anymore on anyone’s current laptop, three slightly different versions of “AS-IS process map v4_FINAL_FINAL_USE-THIS-ONE”, and a single, lonely Word document called Lead-to-Cash Process Overview that was last edited the week before the author left the company. Together with her expertise and knowledge.
I’ve been in and around this “folder” for nearly three decades. I’ve helped fill it, audit it, archaeologically excavate it, and quietly weep over it. Somewhere along the way, I discovered a way of taking notes and a pattern that eventually grew into Universal Process Notation - UPN for short - partly out of conviction that process documentation deserved a better methodology, partly out of the petty satisfaction of wanting to never see a swim-lane spaghetti diagram again…
UPN is brilliantly simple. One symbol, the rectangle. Every activity must answer five fundamental questions: WHAT you do, WHY you do it, WHEN does it start, WHO is responsible, and HOW (which becomes a child diagram when the answer gets interestingly more complex). That’s it. No diamonds, no gateways, no XOR/AND/OR labels that nobody ever reads, let alone understands. A new joiner in marketing can pick up a UPN diagram and understand it. Try doing that with a BPMN export.
But here’s the thing any methodology can’t fix. The methodology is actually mostly the easy part.
The bit that breaks consultants
The actual labour of producing structured process knowledge is brutal.
You start with a discovery workshop. You collect transcripts, interview notes, policy PDFs, three RACI matrices that disagree with each other, a few training decks, screenshots of someone’s whiteboard, and, if you’re really lucky, a SOP from 2018 that was technically accurate at the time. Then you spend the next six weeks turning that pile into something coherent. You translate, structure, draw, redraw, chase down inconsistencies, re-interview the SME because the policy says one thing and the workshop said another, and eventually produce a diagram that lands in a platform like Elements.cloud. And gets reviewed in about ninety minutes.
The diagram is great. The six weeks before the diagram are the problem.
And here’s the perverse part: those six weeks were never the valuable bit. The valuable bit is the validation. The conversation where the people who actually do the work look at the proposed structure and tell you what’s wrong with it. That’s where the insight comes out. That’s where the “wait, we still do that?” moments happen. That’s where process improvement starts.
We’ve been spending 95% of our time on the structuring grunt work to unlock 5% of the time where the actual value gets created.
I have, let us say, a couple of opinions about this.
What changed
For some time now, the temptation to throw AI at this problem was strong, and the results were embarrassing. Early LLMs would happily invent activities, hallucinate resources, and produce diagrams that looked plausible if you didn’t really read them. They captured nothing of the WHY. They couldn’t tell the difference between a business justification and a system status. They made the existing problem worse by giving you confident-sounding nonsense to wade through.
Then, somewhere in the last eighteen months, the models got good enough. Not perfect, but good enough at the specific task of taking messy input and producing structured output that a human can sensibly review. The remaining error rate is exactly the kind a person validating the output catches in minutes, not the kind that contaminates a whole project. With models like Claude Sonnet 4.6, the structuring grunt work moved from “AI can’t be trusted with this” to “AI can do the first pass, and a human can finish.”
That’s the gap Synfliq closes. Not “AI does process mapping for you” (please, no) but “AI does the structuring drudgery, a human does the validation, and the six weeks become six hours.”
So what is the Synfliq proposition, actually?
It is clearly not a process mapping tool. Drawing diagrams is not the bottleneck. For example, Elements.cloud already does that beautifully, and I’m not in the business of building yet another competing canvas.
Synfliq is the layer that lives one step earlier in the workflow. You pour your scattered content in, e.g., DOCX procedures, XLSX RACI matrices, PPTX training decks, PDF compliance manuals, photos of whiteboards, recordings of process interviews, the occasional BPMN export from a system nobody opens anymore… Synfliq turns it into structured, UPN-compliant process knowledge. It generates a parent-child hierarchy. It preserves boundary consistency between levels. It assigns resources, applies RASCI, writes flowline labels that answer business WHY rather than recording past-tense status. It actually produces a package that imports cleanly into Elements.cloud, where you do the part that actually matters: review, amend, approve.
The blunt version of the proposition: scattered content already exists inside your organisation. The structured knowledge it represents is what’s missing. Synfliq’s job is to bridge that gap.
The shift is from we produce documents to we maintain and leverage knowledge in the right context. Documents are artefacts; they go in the folder and decay. Process knowledge is infrastructure; it stays current, traceable, and structured, and other systems and people can rely on it.
Where Synfliq sits, where it doesn’t
There’s a small zoo of AI tools nearby in this space, and people sometimes ask whether Synfliq is competing with them. Mostly it isn’t.
Scribe for example captures what you click. It watches a screen, records the steps, produces a guide. It is excellent at the WHAT and entirely silent on the WHY. It will tell you the user clicks “Approve”. It will not tell you whether they’re approving because of a regulatory requirement or because Kevin needs it before lunch.
Elements.cloud reads your Salesforce metadata and infers the configuration. It is excellent at the HOW of an existing org. It is structurally bottom-up: it sees what your system does, how an org is configured, not what your business is trying to accomplish.
dx0 and friends map also what’s in your Salesforce org. Synfliq maps the processes that use your Salesforce org. Genuinely complementary, when you squint at it the right way.
Synfliq sits therefore in a different lane. We synthesise the WHY out of the content that already documents it; the policies, the transcripts, the procedures, the conversations. And we do it in a format that the rest of the toolchain can consume. We are a “Why” engine, in a market full of “What” tools and “How” tools. And that adds way more value.
On the question of trust
There is a particular kind of AI vendor I have no interest in becoming. The kind that asks you to trust the model.
Synfliq doesn’t ask for that, because the AI is not the authority. You are. Every output Synfliq produces carries its provenance: which source documents fed it, which operation was applied, when, by whom. The customer reviews, amends, approves. The trust comes from the chain: “this output traces to my own content, validated by my own team” and not from us telling you the model is benchmarked.
The implication is liberating, once you sit with it. As Synfliq’s outputs don’t have to be perfect to be useful. They have to be traceable, fast, and accelerative. The human review fills the gap between AI output and verified knowledge. That’s not a workaround for AI’s limitations, it’s the design.
This is also why Synfliq doesn’t require pristine inputs. Process knowledge in real organisations is genuinely messy: incomplete, contradictory, dated, partially documented. A tool that requires clean inputs solves the easy half of the problem. Our job is to make the messy half tractable.
Where this is going
Today, Synfliq primarily accelerates a workflow. You pour content in, you get structured knowledge out, you take it to Elements.cloud, you publish. That’s Phase 1: the accelerator. It’s what’s live at app.synfliq.com right now, and it’s already doing the six-weeks-to-six-hours job for early users.
Phase 2 turns that one-shot generation into something that accumulates. The diagrams you produce become a knowledge vault that survives across projects. Operations apply against prior state: extend, refactor, refine, sync. E.g., your amendments in Elements.cloud flow back. Each iteration enriches what’s there. The longer you use it, the more your organisation’s process knowledge becomes a defensible asset rather than yet another folder full of stale PDFs.
A recent conversation made the demand for this concrete. A team had started experimenting with Claude Code to extract Elements.cloud diagrams to JSON, manipulate them with AI, and re-import the modified version. The instinct is exactly right; as that round-trip is the future of working with structured process knowledge. The hard part is the bit in the middle: keeping the JSON structurally valid through the AI pass, preserving the parent-child boundary consistency, the three-tier ID system, the format Elements.cloud expects on the way back in. Without those guardrails, the round-trip becomes a DIY script with crossed fingers. That’s the work Synfliq does: turning the round-trip from a hopeful experiment into a product feature with provenance attached.
Phase 3 (and here is where I get to be cheerfully ambitious) makes the whole thing ambient. The knowledge surfaces wherever it’s needed: in Salesforce, in Slack, in a Teams notification, in a structured response to an AI agent asking “what are the approval steps in our invoice process?” via MCP. Process knowledge stops being a destination you visit and becomes infrastructure your organisation runs on. Most knowledge workers, in that future, never visit Synfliq at all. That is exactly the point.
A small invitation
If you’ve spent any time in the SharePoint folder I described in the opening paragraph, I’d love to hear from you. If you’re a process consultant who has felt the slow erosion of soul that comes from manually structuring your eleventh discovery output of the quarter: likewise. If you’re a Salesforce practice lead watching six-week discovery cycles eat your margins: definitely.
Synfliq is live. The first wave of early adopters are in. The methodology is right, the technology is getting better every day, the timing is correct, and I have a notable amount of accumulated frustration to convert into product velocity.
There will be more posts. Some will be about the engineering and lessons learned. Some about the methodology, with strong opinions held confidently. Some about partnerships and where this lands in the broader ecosystem. And some, inevitably, about that folder.
— Walter
Synfliq is the structured process knowledge layer for organisations who’ve stopped pretending another round of process documentation will fix it. Try it at app.synfliq.com.