Most AI projects fail not because of the technology but because the vendor never understood the process. So our forward-deployed engineers go on site, sit next to your teams and own the outcome: a solution people use every day, measured against a baseline you sign.
Every engagement has a lead engineer. They run the diagnostic, build the pilot and stay through launch. You never have to re-explain the problem to three different teams, and nobody hands your project over to a delivery center you have never met.
Interviews the people doing the work, watches how it actually gets done, and finds where time is lost.
Connects systems, configures agents, writes integrations. Decisions get made on the spot, without a chain of sign-offs.
By the end, your staff can change agent instructions and add new scenarios themselves.
The real process is never the one in the procedure manual. Two days next to the people who do the work tell an engineer more than two months of requirement documents.
A typical plan for one process and one user group. The diagnostic takes three weeks, the pilot eight. The dated version of this chart is the first page of your pilot plan.
Timelines are typical for a mid-sized engagement. The dated plan is fixed at the end of the diagnostic, and the diagnostic fee is credited to the pilot.
Each phase ends with a document, not a slide show. They are written for the people who approve budgets: short, numbered, and traceable to the data they came from.
Every process we looked at, scored on frequency, data readiness, measurability, ownership and risk. Three to five candidates are ranked, with one recommended for the pilot.
Baseline numbers for the chosen process, a conservative and an expected case, the cost of the pilot and of the first subscription year, and the dated week-by-week plan.
What was built, how it was measured, and what changed against the baseline. It ends with a clear recommendation: scale, adjust or stop, and what scaling would cost.
How the solution runs in your environment: architecture, access model, monitoring, incident procedures and how your team changes agent instructions safely.
Every engagement runs on the same cadence. You always know what was done last week, what happens next, and what decision is waiting for you.
The lead engineer shows working software on your data, not status slides. The process owner and pilot users attend; open questions are logged with an owner and a date.
One page: done, next, risks, decisions needed. Sent the same day as the demo, so the sponsor can follow without attending.
Sponsor, process owner, IT and security, our lead engineer. Reviews progress against the plan, resolves blockers, approves scope changes in writing.
We present the report and a live demo. You sign an acceptance certificate against the criteria agreed at the start, or we fix what is missing first.
After launch, a review of agreed metrics, agent quality and adoption, with the next processes proposed. Frequency depends on the package.
A clear split of responsibilities is what gets a project finished on time.
| Task | BlackGust | Client |
|---|---|---|
| Project goals and success criteria | Proposes and quantifies | Approved by the sponsor |
| Process description | Runs interviews, documents | Process owner, 2–4 h per week |
| Data access | Prepares requirements and schema | IT grants access, security approves |
| Development and configuration | All of it | — |
| Testing | Reference checks | Users test on their own tasks |
| Training | Delivers | Makes staff time available |
| Phase acceptance | Report and demo | Signs the acceptance certificate |
R responsible · A accountable · C consulted · I informed. One accountable party per line.
| Activity | Sponsor | Process owner | IT & security | Pilot users | BlackGust lead | BlackGust team |
|---|---|---|---|---|---|---|
| Diagnostic | ||||||
| Goals and success criteria | A | C | I | — | R | I |
| Interviews and process mapping | I | C | I | C | A | R |
| Data and systems audit | I | C | C | — | A | R |
| Choice of pilot process | A | C | C | — | R | I |
| Pilot and integration | ||||||
| Access to systems and data | I | I | A | — | R | C |
| Security review and approval | I | — | A | — | C | R |
| Building agents and integrations | I | C | C | — | A | R |
| Testing on real tasks | I | A | — | R | C | C |
| Training and documentation | I | C | C | I | A | R |
| Phase acceptance | A | R | C | I | R | I |
We only claim what we can measure. The baseline is taken during the diagnostic, before anything changes, and the same metric is measured the same way after the pilot.
| Metric | Baseline | After pilot | Change |
|---|---|---|---|
| Weekly management report, hours | 6.5 | 0.8 | −88% |
| Counterparty reconciliation, days | 4 | 1 | −75% |
| Requests routed without manual triage | 0% | 71% | +71 pp |
| Answers matching expert reference | — | 94% | target 90% |
| Cases escalated to a human | — | 18% | by design |
Recommendation: scale to two more branches. Payback on the conservative case within the first subscription year.
Adoption is planned like any other workstream. Training is role-based, in your language, on your own data, and it starts during the pilot, not after launch.
How to query the console, read sources, and set up the daily digest. Usually one-to-one, in the executive's own office.
Small groups work through their own cases with the engineer in the room. Short guides and videos stay behind for new staff.
Roles and access, monitoring, editing agent instructions, adding scenarios, and the safe-change procedure.
A trained champion in each team collects feedback, helps colleagues and joins the weekly demo. They are the reason adoption survives our departure.
The method stays the same in every package. What changes is the size of the team, how much of it sits on your site, and how often we review impact together.
| Foundation | Enterprise | Sovereign | |
|---|---|---|---|
| Team | |||
| Forward-deployed engineers | 1 lead engineer | Lead + 2 engineers | Embedded team of 5+ and an architect |
| Presence on site | At key phases: kick-off, diagnostic, launch, acceptance | Continuous, on site and remote | Embedded with your teams |
| Scope | |||
| Business functions | One | 3–5 | Organization-wide |
| Processes | Up to 3 | Up to 15 | Unlimited |
| Connected systems | Up to 5 | Up to 15 | Unlimited |
| Governance | |||
| Impact review | Quarterly | Monthly | Monthly, with the steering committee |
| Support | Business hours | 24/7 for critical incidents | 24/7, 1-hour response, named duty engineer |
Annual platform packages: Foundation $90,000, Enterprise $250,000, Sovereign $490,000 per year. Full comparison on the pricing page.
Our engineering team is based in Tashkent (UTC+5). Engineers travel for the phases that need a room; the remote cadence runs inside your business day.
On site. Kick-off and interviews, pilot launch, acceptance; longer stays during integration.
Remote cadence. We are three to four hours ahead of Central Europe, so our day covers your morning and early afternoon. Weekly demos and steering meetings sit in that window.
On site. Diagnostic interviews and launch weeks on site; Enterprise and Sovereign teams stay longer.
Remote cadence. Remote engineers shift their hours to give a daily overlap with your morning. Demos and steering meetings are booked in your business hours.
On site. Kick-off, diagnostic and acceptance on site; air-gapped work is done on your premises.
Remote cadence. We are three to six hours behind East Asia and Australia, so your afternoon is our morning. Work runs in your language, with interpreters arranged where needed.
We have seen the same handful of risks in most projects. Each one has an owner and a mitigation in the pilot plan from day one.
We handle the technology. The organization provides decisions and access; without them no vendor can ship a working product.
A software engineer who works inside your organization, next to the people who do the work. They map the process, connect systems, configure agents and stay accountable for the result. It is the opposite of a consultant who writes a report and a separate team that builds something else.
Usually about eleven weeks: a diagnostic of 2–3 weeks and a pilot of 4–8 weeks. You see working software on your own data at the weekly demos from around week five.
The diagnostic costs $1,490 and is credited to the pilot if you continue. If you stop, you keep the opportunity map and ROI estimate. They are written so another vendor or your own team could act on them.
The process owner spends 2–4 hours a week, IT a few hours to grant access, and the pilot group of 5–20 users works with the solution on real tasks in the second half of the pilot. The sponsor attends the steering committee every 2–4 weeks.
No. They are on site for kick-off, interviews, launch and acceptance, and for longer periods in Enterprise and Sovereign engagements. Air-gapped work is always done on your premises. The rest runs remotely in your business hours.
Yes. Engineers travel to the client for on-site phases, and the remote cadence is scheduled inside your working day. Deployment can stay in your region or on your own infrastructure to meet data-residency rules; see security.
By measuring a baseline before anything changes and the same metric after launch, with the method written down at the start. The process owner co-signs the numbers in the pilot report.
Your data, agent instructions, process documentation and reports stay yours. The platform is licensed through an annual package; details are in the contract and on the pricing page.
If the pilot meets its acceptance criteria, integration takes 2–4 months within the first subscription year: all required systems, single sign-on, training and handover to SLA support. Then we add processes one by one.
The data model, modules and agents behind every engagement.
Open →SolutionsWhere pilots usually start, and the metric each one moves.
Open →PricingFoundation, Enterprise and Sovereign, side by side.
Open →SecurityCloud in your region, private cloud, on-premises or air-gapped.
Open →The first consultation is free. Tell us about the problem and an engineer will tell you where to start.