نقشه‌ی خروج از وابستگی به وندور وقتی طرف ورشکست می‌کنهEXCERPT: وقتی وندوری که بهش وابسته‌ای علائم مشکل مالی نشون می‌ده یا مستقیم اعلام ورشکستگی می‌کنه، این Prompt وحشت رو تبدیل می‌کنه به یه نقشه‌ی مرحله‌ای: میزان قرار گرفتن در معرض ریسک، آبشار خرابی، تایم‌لاین و گزینه‌های جایگزین رتبه‌بندی‌شده.
چرا این پرامپت اهمیت دارد
Teams that start vendor-failure planning only after the bankruptcy headline hits typically burn through half their actual runway just figuring out what's integrated where — by the time they're ready to migrate, the data-export window has already shrunk. Rush migrations run 2-3x the cost of planned ones, and companies hit by a sudden vendor shutdown routinely lose some customer data entirely because nobody exported it before access was cut.
ما از آن برای چه استفاده می‌کنیم
Your company's AI-powered customer support chatbot vendor just filed for Chapter 11 bankruptcy — similar to Character.AI's October 2026 filing — and legal has told you the data-export window closes when the court-approved sale concludes, in 6 to 8 weeks. You have no idea whether your team can actually migrate in that window, what breaks first, or what to tell leadership at Friday's update.
پرامپت
ROLE You are a vendor risk management consultant who has guided companies through dozens of vendor failures, bankruptcies, and forced migrations. You think in terms of blast radius, switching costs, and time-to-failure — not vendor marketing claims. CONTEXT My company depends on [VENDOR NAME] for [SPECIFIC FUNCTION, e.g. "customer support ticketing," "LLM inference for our chatbot feature," "payment processing"]. Here is what I know about the vendor's situation: [PASTE ANY SIGNALS: layoffs, bankruptcy filing, missed SLAs, executive departures, funding news, customer complaints, etc.] Here is how deeply integrated this vendor is into our systems: [DESCRIBE: e.g. "Direct API calls from 12 microservices," "Embedded widget on our checkout page," "Core data store with 2 years of customer history"] Our contractual situation: [DESCRIBE: contract length remaining, data export clauses, SLA guarantees, termination notice period — or write "unknown, need to check"] TASK Produce a risk-ranked contingency plan that answers: 1. What is our actual exposure if this vendor disappears with [NO NOTICE / 30 DAYS NOTICE / 90 DAYS NOTICE] (pick the most likely scenario based on what I told you about their situation)? 2. What data or functionality would we lose access to first, and what breaks downstream? 3. What is the realistic migration timeline to a replacement, broken into: emergency stopgap (get something working in days), interim bridge (weeks), and proper migration (months)? 4. What should we do THIS WEEK regardless of whether the vendor actually fails — i.e., what are the no-regret moves? 5. What are 2-3 alternative vendors or approaches, ranked by migration friction (not by feature comparison)? CONSTRAINTS - Do not assume best-case scenarios. Assume the vendor's wind-down process will be slower and messier than their official communications suggest. - Flag any point where you are making an assumption because I didn't give you enough information, rather than silently guessing. - Do not recommend "wait and see" as a standalone strategy — every plan must include at least one action to take immediately. - Be specific about dependencies: if migrating away from this vendor requires migrating something else first, say so explicitly. OUTPUT FORMAT Structure your response as: 1. Exposure Summary (2-3 sentences: how bad is this, really) 2. Failure Cascade (bulleted list: what breaks, in what order, if the vendor disappears) 3. Contingency Timeline (table or structured list: Emergency / Bridge / Full Migration, with rough time estimates for each) 4. This Week's Actions (numbered list, 3-5 items, no-regret moves) 5. Alternative Vendors/Approaches (ranked list with one-line rationale for the ranking)
نتیجه
Exposure Summary
Your support chatbot handles roughly 40% of first-contact tickets today, which means losing it doesn't crash your support queue — it roughly doubles your human agents' workload overnight. The real risk isn't the chatbot itself; it's the 14 months of resolved-ticket history stored only in the vendor's system, which you're using for quarterly QA reviews and new-agent training. If the data-export window closes before you pull that archive, you lose your only record of how disputes were actually resolved.
Failure Cascade
- Chatbot widget stops responding on your support page (immediate, cosmetic — customers see a broken widget, not missing support)
- Ticket auto-routing rules that reference the vendor's tagging system break, so new tickets land in a general queue instead of being pre-sorted by issue type
- Your QA dashboard, which pulls resolution-time metrics from the vendor's API, goes blank — you lose visibility into whether support SLAs are being met
- New-agent onboarding materials that link to "example resolved tickets" in the vendor's archive become dead links
- If the vendor's wind-down includes database deletion (not just service shutdown), the 14-month ticket history disappears permanently
Contingency Timeline
- Emergency stopgap (3-5 days): Disable the broken widget, route 100% of chat volume to human agents, communicate the temporary change to the support team
- Bridge (2-3 weeks): Stand up a basic FAQ deflection page or a lightweight open-source chatbot to handle the simplest 20% of tickets while you evaluate replacements
- Full migration (6-10 weeks): Select and implement a replacement platform, re-train routing rules, rebuild the QA dashboard integration
This Week's Actions
- Export all 14 months of ticket history today — don't wait for a "final warning" from the vendor, because bankruptcy wind-downs rarely give one
- Identify every system (routing rules, QA dashboard, onboarding docs) that references the vendor's API or tagging taxonomy, not just the visible chat widget
- Get written confirmation from the vendor (even a form response) of the exact date data export access ends
- Brief your support team lead that manually-handled ticket volume will roughly double — adjust staffing or triage rules now, not after the widget goes dark
- Start a shortlist of 2-3 replacement vendors so you're not starting vendor evaluation from zero once urgency hits
Alternative Vendors/Approaches (ranked by migration friction, not features)
- Intercom or Zendesk's native AI add-on — lowest friction if you already use either as your base ticketing system; the chatbot layer bolts onto infrastructure you're not replacing
- Open-source RAG chatbot on your own ticket history — medium friction; more engineering work upfront, but you own the data and avoid repeating this exact problem with a new vendor
- Pure human staffing increase, no chatbot replacement — lowest technical friction but highest cost; often the right call for a 60-90 day bridge even if you plan to replace the chatbot eventually
شکست یک تأمین‌کننده از آن ریسک‌هایی هست که همه در سطح کلی قبولش دارن و تقریباً هیچ‌کس براش برنامه‌ی مشخص نداره — تا وقتی یه تیتر مثل پرونده‌ی Chapter 11 شرکت Character.AI در اکتبر ۲۰۲۶ صبح یه سه‌شنبه برسه دستت و یکی بپرسه «ما هم از چیزی شبیه این استفاده می‌کنیم؟» این Prompt ساخته شده تا همون لحظه‌ی وحشت رو در یک پاس به یه برنامه‌ی ساختاریافته تبدیل کنه.
چرا نقش اینجا مهمه
چارچوب نقش — «مشاور ریسک تأمین‌کننده»، نه «دستیار کمک‌کننده» — برای خنثی کردن یه الگوی خطای مشخص هست: مدل‌های AI وقتی درباره‌ی ریسک تأمین‌کننده ازشون پرسیده می‌شه، معمولاً به دلجویی می‌افتن («احتمالاً مشکلی نیست، بیشتر شرکت‌ها از این وضعیت بیرون می‌آن»). چارچوب مشاور، مدل رو به سمت فکر کردن به شعاع تخریب (blast radius) می‌بره، نه به سمت آروم‌کردن خیال.
چرا ورودی زمینه به سه بخش تقسیم شده
بیشتر آدم‌ها وقتی یه مشکل تأمین‌کننده رو توضیح می‌دن، توضیح می‌دن که اون تأمین‌کننده چه کاری می‌کنه و همون‌جا متوقف می‌شن. این Prompt سه ورودی جدا رو الزام می‌کنه — وضعیت خود تأمین‌کننده، عمق یکپارچه‌سازی، و شرایط قرارداد — چون هر کدوم به‌طور مستقل جواب رو تغییر می‌ده. تأمین‌کننده‌ای که ۹۰ روز مهلت اطلاع‌رسانی قراردادی داره ولی یکپارچه‌سازی فنی عمیقی داره، به برنامه‌ای متفاوت از تأمین‌کننده‌ای نیاز داره که هیچ تضمین مهلتی نداره ولی یکپارچه‌سازیش سطحیه. چپوندن اینا توی یه توده‌ی واحد، همین تفاوت رو از بین می‌بره.
چرا تایم‌لاین سه فازه، نه یکی
یه «برنامه‌ی مهاجرت» واحد، فلج تحلیلی به بار می‌آره — تیم‌ها هفته‌ها صرف ارزیابی جایگزین بی‌نقص می‌کنن در حالی که سرویس واقعی داره افت می‌کنه. تقسیم به سه فاز — راه‌حل موقت اضطراری، مرحله‌ی گذار، و مهاجرت کامل — به تیم اجازه می‌ده ظرف چند روز دست به اقدام بزنه و همزمان به سمت یه راه‌حل واقعی حرکت کنه، نه اینکه تا پیدا شدن جواب «درست» منجمد بشه.
چرا جایگزین‌ها بر اساس اصطکاک رتبه‌بندی می‌شن، نه ویژگی
مقایسه‌ی ویژگی‌ها کاریه که وقتی وقت داری انجام می‌دی. تحت فشار شکست تأمین‌کننده، سوال این نیست «کدوم ابزار بهتره» — سوال اینه «کدوم ابزار رو واقعاً می‌تونیم راه بندازیم قبل از اینکه ابزار فعلی ناپدید بشه.» رتبه‌بندی بر اساس اصطکاک مهاجرت، گزینه‌ی کسل‌کننده-ولی-سریع رو بیرون می‌کشه که یه مقایسه‌ی ویژگی-محور دفنش می‌کرد.
مهم‌ترین محدودیت
«به‌عنوان استراتژی مستقل، صبر-و-ببین رو توصیه نکن» مهم‌ترین جمله‌ی این پرامپت است. مدل‌ها (و آدم‌ها) وقتی ورودی مبهمه، به «وضعیت رو رصد کن» می‌افتن. الزام به حداقل یه اقدام فوری، هر بار، همون چیزیه که این رو از یه به‌روزرسانی وضعیت به یه برنامه‌ی اقتضایی تبدیل می‌کنه.